Pacic-Basin Finance Journal 30 (2014) 4461

Contents lists available at ScienceDirect

Pacic-Basin Finance Journal

journal homepage: www.elsevier.com/locate/pacfin

How protable is the Indian stock market?

Paresh Kumar Narayan a,, Huson Ali Ahmed a,
Susan Sunila Sharma a, Prabheesh K.P. b
a
b

Financial Econometrics Group, School of Accounting, Economics and Finance, Deakin University, Australia
Department of Liberal Arts, Indian Institute of Technology, Hyderabad, India

a r t i c l e

i n f o

Article history:
Received 17 September 2013
Accepted 4 July 2014
Available online 11 July 2014
JEL classication:
G11
G12
G14
G17

a b s t r a c t
In this paper, using a range of technical trading and momentum
trading strategies, we show that the Indian stock market is protable.
We nd robust evidence that investing in some sectors is relatively
more protable than investing in others. We show that sectoral
heterogeneity with respect to protability is a result of the gradual
diffusion of information from the market to the sectors. Specically,
we show that while the market predicts returns of sectors, the
magnitude of predictability varies with sectors. Our results are robust
to a range of trading strategies.
2014 Elsevier B.V. All rights reserved.

Keywords:
Momentum
Technical trading
Prots
Sectors
Stock market
India
Predictability

1. Introduction
The growth of the Indian stock market over the last decade has been impressive, particularly in terms
of market capitalization, number of listed companies, and turnover rate. Market capitalization as a
percentage of GDP of the National Stock Exchange (NSE), for instance, increased from 35% in 2001 to 85%
in 2011. Similarly, the number of companies listed on the NSE more than doubled over the corresponding
period, from 720 to 1552. Likewise, the turnover on the Indian stock market increased from US$621 billion
Corresponding author at: School of Accounting, Economics and Finance, Faculty of Business and Law, Deakin University, 221 Burwood
Highway, Burwood, Victoria 3125, Australia. Tel.: +61 3 9244 6180; fax: +61 3 9244 6034.
E-mail addresses: paresh.narayan@deakin.edu.au (P.K. Narayan)., huson.aliahmed@deakin.edu.au (H.A. Ahmed).,
s.sharma@deakin.edu.au (S.S. Sharma)., Prabheesh@iith.ac.in (P. K.P.).

http://dx.doi.org/10.1016/j.pacn.2014.07.001
0927-538X/ 2014 Elsevier B.V. All rights reserved.

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

45

in 2010 to US$1056 billion in 2011. For all its impressive growth, India achieved a global rank of 7th in
terms of market capitalization, 10th in terms of total value traded, and 22nd in terms of turnover ratio, as
of December 2010 (NSE, 2011).
The impressive growth of the Indian stock market has attracted research on the efciency (or
otherwise) of the market. There are a number of studies that examine the efciency of the Indian stock
market (see, inter alia, Dicle et al., 2010; Kumar et al., 2011; Majumder, 2013; Mishra et al., 2011; Narayan
et al., 2014a,b; and Narayan and Ahmed, in press). There are two ways in which this literature can be taken
forward. First, these studies do not provide an economic signicance analysis with respect to market
efciency or inefciency. Therefore, it is unclear how protable the Indian stock market is. Second, if an
investor wants to invest in the Indian stock market, which sectors should she invest in? This question
remains unanswered in these studies.1
Therefore, from this growing literature on the Indian stock market, there are clearly four things which
are relatively less understood. These are:
1. The Indian stock market may be protable but do prots vary from sector-to-sector?
2. Can investors use different trading strategies to make prots from the Indian stock market? If yes, how
much do the prots vary from sector-to-sector as trading strategies change?
3. Short-selling is a feature of the Indian market; therefore, what effect does short-selling has on sectoral
prots?
4. The most recent global nancial crisis has affected stock markets globally; therefore, has the crisis
affected sectoral prots on the Indian stock market?
To the best of our knowledge, none of these questions has been answered in the literature. Our study
provides the rst attempt at addressing each of these questions. Our approach is as follows. We utilise a
range of technical trading rules and momentum trading rules to identify winners and losers and form
returns for momentum and zero-cost portfolios. We also consider ranking stocks based on moving average
rules and then undertaking long and short positions. Taking long and short positions based on a ranking of
sectors allows an investor to diversify risk and consider a portfolio of sectors in her trading strategy. More
details on our approach are provided in the next section.

2. Approach
2.1. Moving average ranking-based trading strategy
In this section, we consider trading strategies based on moving average rules. The moving average
technical trading rules are popular trading strategies, particularly in the foreign exchange market (see, for
instance, Lee and Mathur, 1996a,b, and Szakmary and Mathur, 1997) and commodity markets (see
Narayan et al., 2013; Narayan et al., 2014b). We begin as follows:
1. We compute the monthly long-run (LR) and short-run (SR) moving averages (MA) using identied
intervals for LR and SR.
2. Using (SRLR) moving average for respective intervals, we rank each of the six sectors, from best (rank
1) to worst (rank 6). The highest positive difference between SRMA and LRMA is assigned a rank of 1,
and the lowest difference between SRMA and LRMA is ranked last.
3. We initiate long positions in high ranked sector(s) and short the sector with the lowest rank. In this
way, we are able to devise three trading strategies (excluding the nave investor strategy), as will be
discussed soon.
4. When we do not allow for short-selling, we simply do not short but allow for cash positions.
1
There is another branch of this literature (Chang et al., 2004), which typically employs simple technical trading rules, such as the
moving average and the trading range break rules. These studies generally nd some evidence of protability. Similarly,
Gunasekarage and Power (2001) and Narayan et al. (2014a) nd evidence that the Indian stock market is protable, although these
studies use different approaches compared to what we do in this paper.

46

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

Following the Lee and Mathur (1996a,b) and the Okunev and White (2003) studies, we dene the
short-run moving average (MASR) and the long-run moving average (MALR) at time t as:
SR

Rt j1MASR
j;t1
j
LR
Rt i1MAi;t1

MA j;t

LR
MAi;t

where j and i are prior months of sectoral returns. The SRMA values range from 1 to 12 months, while the LRMA
values range from 2 to 36 months. In all combinations of short-run and long-run moving average trading rules,
j b i. Just to demonstrate how these trading rule combinations are implemented, consider the sectoral position
LR
SR
LR
SR
LR
based on a SRMA of 1 month. We compute this as MASR
1t MA2,t, MA1,t MA3,t, , MA1,t MA36,t. Similarly,
SR
SR
LR
SR
LR
using a 2-month short-term moving average MA2,t, MA2,t MA3,t, , MA2,t MA36,t, we determine the
investment positions. At the end of each month, for each individual moving average combination, the six sectors
are ranked on a scale of 1 (best-performing) to 6 (worst-performing). This ranking is based on the return-based
momentum indicator, which is simply the difference in return between the SR and LR moving averages. The
Table 1
A summary of trading rules and trading strategies.
Each month, from January 2001 through December 2012, each sector is ranked from one to six based on the difference between the
short-run MA and long-run MA of prior returns, using MA combinations for each strategy under the respective scenarios. Using the
differences between the short-run and long-run MA, we take a long position in the most attractive sector (rank 1) and a short
position in the least attractive sector (rank 6) in the case of short-selling, or simply take a cash position in the case of no short-selling.
The short-run MA parameter ranges from 1 month to 12 months, while the long-run MA parameter runs from 2 months to
36 months.
Strategy

MA rules (in number of months)

Panel A: scenario I: with 100% short-selling allowed

Two
[1,2][12,36]

Three

[1,2][12,36]

Four

[1,2][12,36]

Panel B: scenario II: short-selling not allowed

One: naive strategy
[1,2][12,36]
Two
[1,2][12,36]
Three

[1,2][12,36]

Four

[1,2][12,36]

Panel C: scenario III: trading strategies based on limited short-selling

Two
[1,2][12,36]

Three

[1,2][12,36]

Four

[1,2][12,36]

Long and short positions

Long position: rank 1 (100%)
Cash position: rank 2,3,4,5 (0%)
Short position: rank 6 (100%)
Long position: rank 1 (50%)
Long position: rank 2,3 (25%)
Cash position: rank 4,5 (0%)
Short position: rank 6 (100%)
Long position: rank 1 (70%)
Long position: rank 2 (30%)
Cash position: rank 3 (0%)
Short position: rank 4 (100%)

Invest equally across 6 sectors

Long position: rank 1 (100%)
Take no position: rank 2,3,4,5,6 (0%)
Long position: rank 1 (50%)
Long position: rank 2,3 (25% each)
Take no position: rank 4,5,6 (0%)
Long position: rank 1 (70%)
Long position: rank (30%)

Long position: rank 1 (100%)

Long position: rank 2 (76%)
Take no position: rank 2,3,4,5, (0%)
Short position: rank 6 (76%)
Long position: rank 1 (76%)
Long position: rank 2,3,4,5 (25% each)
Short position: rank 6 (76%)
Long position: rank 1 (70%)
Long position: rank 2 (30%)
Long position: rank 3,4 (38% each)
Short Position: rank 6 (76%)

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

47

sector that has the largest return difference (MASR MALR), earns a rank of 1, the sector with the second largest
return difference earns a rank of 2, while the sector with the least return difference is the worst-performing and,
thus, earns a rank of 6. We place weights on each of the short-run/long-run combinations and positions are taken
and held for a month, based on the respective strategy. The rankings are revised each month and new positions
are taken if rankings do indeed change.
We propose four different strategies to examine the efciency and protability of the sectors in the
Indian market. A summary of the trading strategies is provided in Table 1.
Strategy I: This strategy assumes a nave investor who invests equally in all six sectors, irrespective of
the sectors' rankings.
Strategy II: This strategy invests 100% in the highest ranked sector and shorts the lowest ranked sector in
the case of short-selling, but takes cash positions in the case of no short-selling. This strategy,
thus, provides opportunities to make arbitrage prots from taking two opposite positions.
Strategy III: This strategy invests in the top-3 ranked sectors. An investor with this strategy invests 50% in
the sector ranked one, and 25% each in the sectors ranked two and three. The investor will
short the worst-ranked sector in the case of short-selling and take cash positions in the case
of no short-selling. This strategy, therefore, allows an investor to build a portfolio of sectors
which also diversies risks.
Strategy IV: This strategy allows an investor to invest 70% in the highest ranked sector and 30% in the
second-highest ranked sector. The investor continues to short the weakest ranked sector in
the case of short-selling, and takes cash positions in the case of no short-selling.
In all these strategies, where we allow for short-selling, our approach proceeds as follows. First, we
allow for 100% short-selling. Then, in a robustness test of protability, we allow for a short-selling of 76%
based on our estimate of the magnitude of short-selling in the Indian market.

2.2. Technical trading based rules

In this section, we discuss three specic technical trading rules that we utilise to generate buy and sell
signals.2 These are the moving average rule, momentum trading rule, and a lter-based trading rule.
Essentially, we generate prots using these buy and sell signals from each of these three trading rules and
take the average of these prots to make a judgement on the protability of sectors.
The short-run (j) moving average values range from one to three months, while the long-run (i) moving
average values range from nine to 12 months. In all combinations of short-run and long-run moving average
trading rules, j b i. These combinations are as follows: (SRMA1-LRMA9), (SRMA1-LRMA12); (SRMA2-LRMA9),
(SRMA2-LRMA12); and (SRMA3-LRMA9), (SRMA3-LRMA12). From these rules, we generate a buy signal,
denoted by a value of 1 in a particular month t, if SRMA1 N LRMA9, while a sell signal is generated if
SRMA1 b LRMA9. These buy and sell signals are generated for all six combinations of SR and LR moving averages.
Based on these signals, each month, we simultaneously long in the winner (that is, when SRMA N LRMA) and
short in the loser.
We also use momentum trading rules to generate buy and sell signals. With momentum trading rules,
we use two periods (m = 9, 12). Trading signals are generated based on whether or not the current stock
price is higher than the price m months ago: if it is, a buy signal is generated (with a value of 1);
otherwise, a sell signal (with a value of 0) is generated. Based on this rule, we simultaneously long in
winner and short in losers within an industry/sector.
Finally, we use a lter of 10% price threshold as a trading rule. If current return is 10% higher than
average return a buy signal is generated or else the signal is to sell. Accordingly, a long position in winner
and short position in loser stocks are taken.
Every time we take positions, we keep the investment for 1-month, 3-month and 6-month holding
periods. Using these three trading rules, we form average time-series prots for winner, loser, momentum
(winner loser) and zero-cost (winner + loser) portfolios. For each of these portfolios, we compute
2

See Neely et al. (2013) for an example of using technical indicators to predict equity risk premium.

48

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

mean monthly average returns and standard deviation, which is what we report in Tables presented in the
next section.
3. Data and results
3.1. Data
Our data set comes from the National Stock Exchange (NSE) of India. We use the monthly S&P CNX
Nifty, the index of the NSE, which is comprised of 50 stocks from 24 sectors of the Indian economy; for a
nice recent discussion on the Indian stock market, see Narayan et al. (2014a,b). We use only six sectors in
our analysis. We choose these six sectors because they are the only sectors for which data are available
from January 2001 to December 2012. For the rest of the sectors, data start much later, which is not helpful
for our empirical design. This sample size, constrained by the start date is noted in Narayan et al. (2014b).
The main implication here is that at the sector-level there is no historical time-series data for India that
goes beyond 2000; therefore, we, like the empirical literature, utilise what is available. Because our
approach requires a common start date and needs to have as long a time-series as possible, we end up
with only six sectors. These sectors together constitute approximately 60% of total market capitalization,
led by the banking sector (15.3%), energy sector (11.1%), and the FMCG sector (10.2%). The sectoral indices
are as follows:
1. CNX Bank Index: this index is comprised of the 12 most liquid and biggest capitalized Indian banking
stocks.
2. CNX MNC Index: this index is comprised of 15 listed companies in which foreign shareholding is over
50% and/or the management control is vested in the foreign company.
3. CNX IT Index: this index is comprised of 20 Information Technology (IT) stocks, such as IT-related
activities, in particular, IT Infrastructure, IT Education and Software Training, Telecommunication
Services and Networking Infrastructure, Software Development, Hardware Manufacturers, Vending,
and Support and Maintenance.
4. CNX FMCG Index: this index is comprised of 15 stocks from the Fast Moving Consumer Goods industry
(FMCG).
5. CNX Pharmaceutical Index: this is the index of the Pharmaceuticals sector and is comprised of 10
stocks.
6. CNX Energy Index: this is the index of Petroleum, Gas and Power and is comprised of 10 stocks.
A brief discussion of the data is in order here. We report descriptive statistics on returns in Table 2. This
table reports descriptive statistics relating to market (CNX Nifty) and sectoral returns. Returns are
computed as (Pt/Pt 1 1) * 100, where Pt is the price index. Mean returns, standard deviation,
skewness, and kurtosis statistics' are reported. These statistics are based on monthly data for the period
January 2001 to December 2012. The descriptive statistics of the data suggest that monthly market returns

Table 2
Descriptive Statistics.
This table reports the descriptive statistics relating to market (CNX Nifty) and sectoral returns. Returns are computed as (Pt/Pt 1 1) * 100,
where P is the price index. Mean returns, standard deviation, skewness, and kurtosis statistics are reported. We also report the coefcient from
the AR(1) model of returns and IT t-statistic in the parenthesis (see Column III). *** and * denote statistical signicance at the 10% and 1% levels,
respectively.

S&P CNX Nifty

Banking
IT
FMCG
Pharmaceutical
Energy
MNC

Mean

AR(1)

1.2554
2.0732
0.1647
1.3337
1.4228
1.6408
1.1504

0.3516***
0.3021***
0.2221***
0.2343***
0.3193***
0.2972***
0.3333***

(4.4445)
(0.3753)
(2.6948)
(2.8501)
(0.0001)
(3.7316)
(0.0001)

JB_Q-stat
(24)

Std. dev.

Skewness

Kurtosis

36.183* (0.053)
34.713* (0.073)
34.2* (0.081)
20.47 (0.670)
34.099* (0.083)
31.494 (0.140)
34.32* (0.079)

6.3394
8.3150
11.6485
5.0056
5.4157
7.0763
5.6119

0.4282
0.1257
3.2839
0.4906
0.5877
0.0413
0.4685

4.4843
4.0234
26.2201
3.4735
4.4917
4.5763
4.3218

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

49

over the sample period were 1.26% and four sectors (banking, pharmaceutical, FMCG, and energy) had
average returns in excess of market returns. The monthly sectoral returns are in the 0.162.07% range. The
standard deviation reveals that the IT sector is most volatile; it is almost twice as volatile as the market.
The rest of the sectors are less volatile. In fact, three of the six sectors (FMCG, pharmaceutical, and MNC)
are less volatile than the market. Like mean returns, volatility also varies sector-by-sector. All sectors have
a negative skewness except for the banking sector returns. The IT sector returns are the most negatively
skewed. A similar picture emerges when reading the kurtosis statistics, suggesting that the return
distributions are all leptokurtic and more so for the IT sector. Clearly, then, sectoral return behaviour is
different and the IT sector stands in sharp contrast to others.

3.2. Results from momentum-based trading strategies

3.2.1. Results without short-selling
The main ndings on the performance of each of the trading strategies for each of the six sectors are
reported in Table 3. Results in panel A are based on a strategy whereby the investor is assumed to be nave
and, therefore, invests equally in all six sectors, irrespective of their ranking. We call this strategy I. In
panel B, we have an investor who invests 100% in the rst-ranked sector and has no interest in investing in
the other sectors. We call this strategy II. Panel C contains results from a strategy which invests 50% in the
rst-ranked sector and 25% each in the sectors ranked two and three, leaving the rest of the
worst-performing sectors in cash positions. We call this strategy III. The fourth panel consists of returns
from a strategy that invests 70% and 30% in sectors ranked one and two, respectively. Again, as in previous
strategies, the investor leaves the rest of the sectors in cash positions. We call this strategy IV.
Table 3
Performance of long and short strategies with short-sell restriction (short-selling not allowed).
This table reports monthly mean returns without short-selling. The row with Prob N 0 denotes the percentage to total months in
which prots from a given strategy exceeds zero. The row with Prob N market represents the percentage of total months in which
prots from a given strategy exceeds the market return. Here, we use the S&P CNX Nifty as a benchmark for the market. The paired
t-test indicates whether or not the return from each of the strategies is signicantly greater than the market return. *** and **
represent statistical signicance at the 1%, 5% and 10% levels, respectively.
Banking

Energy

IT

MNC

FMCG

Pharmaceutical

Panel A: strategy I
Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N market (%)
Paired t-test

0.303**
1.459
57.273
34.545
1.931

0.164
1.107
61.818
30.909
2.211

0.175
1.150
58.182
33.636
2.187

0.225**
0.971
61.818
31.818
2.109

0.294***
0.852
68.182
35.255
1.989

0.217**
0.904
60.000
32.727
2.2128

Panel B: strategy II
Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N market (%)
Paired t-test

0.286
2.871
9.091
34.545
1.903

0.628**
2.873
8.182
37.273
1.23

0.815**
3.617
20.000
40.909
0.85

0.373**
1.929
3.636
35.455
1.735

0.587***
2.168
20.000
44.545
1.342

0.215
2.536
6.364
33.636
2.661

Panel C: strategy III

Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N market (%)
Paired t-test

0.297
1.971
27.27
36.36
1.942

0.379**
1.701
21.82
33.64
1.745

0.418**
2.095
33.64
38.18
1.640

0.377***
1.291
35.45
31.82
1.783

0.447***
1.447
46.36
37.27
1.633

0.010
1.408
26.36
31.82
2.461

Panel D: strategy IV
Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N market (%)
Paired t-test

0.283
2.225
20.00
35.45
1.887

0.531***
2.192
14.55
35.45
1.462

0.618**
2.696
30.00
39.09
1.27

0.390***
1.481
16.36
32.73
1.772

0.556***
1.730
34.55
37.27
1.456

0.129
1.845
12.73
30.91
2.656

50

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

Evidence obtained from strategy I, where a nave investor invests equally across all six sectors, suggests
that economic prots are statistically insignicant in energy and IT sectors only. Monthly prots for the
remaining four sectors are in the [0.22, 0.30] percentage range. When we look at the evidence from
strategy II, in all sectors, except banking and pharmaceutical, prots are statistically different from zero. In
other words, prots are statistically signicant at the 10% level or better for energy, IT, MNC, and FMCG
sectors. However, the magnitude of prots varies from sector-to-sector. It is the IT sector which gives
investors the most prots with a monthly average return of 0.815%, while the pharmaceutical sector gives
investors the least prot with a monthly average loss of 0.215%. When we collect evidence from strategies
III and IV, we discover protability results consistent with strategy II in terms of the most protable and
least protable sectors. The most protable sector consistently appears to be IT while the least protable
sector is pharmaceutical. When we consider all sectors as one portfolio, the monthly returns from
strategies I to IV fall in the [1.4%, 2.5%] range.
The main implication of our ndings here is that prots across the three strategies, excluding a strategy
that assumes a nave investor, are generally consistent on two fronts: rst, prots are statistically signicant at
the 10% level or better for four of the six sectors, although prots are clearly sector-dependentthat is, the
magnitude of prots is different in different sectors. Second, the IT sector consistently out-performs the other
sectors, while the pharmaceutical sector produces the lowest prots. On the whole, then, the protable
sectors are IT, energy, FMCG, and MNC.
There are other interesting features of our results which are worth highlighting. Generally, we notice
that with higher prots come higher risk. An interesting trend here is that as investors move from a
strategy where they take a long position in the most attractive sector (strategy II) to taking long positions
in more than one sector (strategies III and IV) as part of investment diversication, a decline in risk is
experienced. The standard deviation for the IT sector, for instance, declines from 3.6 in the case of strategy
II to 2.1 and 2.7, respectively, in the case of strategies III and IV. When we consider the performance of the
other ve sectors, we notice a similar trend in that there is a decline in the standard deviation as an
investor moves from strategy II to strategy IV. The implication here is that a strategy (such as our strategy
III), that allows an investor to take a long position in multiple sectors, leads to diversication of risk
although the prots are smaller from such a strategy compared to strategies II and IV.
For each of the strategies, we report some additional statistics to gauge the performance of the
strategies. In particular, we report the probability that mean returns are greater than zero and greater than
the market (Nifty) return. Ignoring the nave investor (Strategy I), we notice that in strategy III, where we
allow an investor to diversify risk by taking long positions in the top-3 sectors, the probability that returns
are greater than zero is highest for all sectors. For example, with strategy III, 34% of the time IT sector
returns are greater than zero. Moreover, when we compare sectoral returns with the market return, we
nd that all three strategies (strategies II, III, and IV) reveal that for the IT, FMCG, and banking sectors, over
35% of the time returns exceed the market return. Finally, the paired t-test reveals that for most sectors
returns from each strategy are statistically signicant relative to the market, mostly at the 10% level of
signicance.
3.2.2. Results with short-selling
3.2.2.1. Background. Short-selling is the practice of selling shares that the seller does not own at the time of
trading. In India, the Securities and Exchange Board of India (SEBI), the regulator of the Indian capital
market, denes a short sale as selling of the shares without having the physical possession of the shares
unless it is either for squaring-up of an earlier purchase in the same settlement of the same stock
exchange, or against the pending deliveries from the same stock exchange pertaining to previous
settlements (SEBI, 1996). There are various phases associated with short-selling activities in the Indian
market. First, in the 1990s, only retail investors were allowed to short-sell. The Badla system and the
weekly settlement system were in place Weekly settlement with T + 5 days ensured that investors made
payments and took delivery of securities within ve days after the day of the transaction. The Badla
system, by comparison, was useful when investors preferred to postpone the settlement of a transaction
from one settlement period to another. Both these systems governing short-selling activities were
instrumental in increasing the volume of trading, which was responsible for high stock price volatility.
This needed a response; the SEBI temporarily banned short-selling in 1998 and 2001. Second, a more

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

51

stringent measure was taken by SEBI, who abolished the Badla system on 2 July, 2001. In fact, the Badla
system was replaced by a rolling settlement system, with T + 3 days settlement system. Third, from 1
April, 2003, the Indian stock market moved to the T + 2 rolling settlement system. Therefore, the
payment and delivery of securities had to be completed within two days after the day of the transaction
(Pathak, 2008).
Fourth, in 2007, the SEBI allowed all classes of investors, including institutional investors, to short-sell
in the Indian stock market. However, naked short-selling was not permitted; all investors are required to
mandatorily honour their obligation of delivering the securities at the time of settlement. Similarly, in
2007, the SEBI introduced a fully-edged Securities Lending and Borrowing Scheme (SLBS) for the
effective transaction of short-selling. Through this scheme, investors can borrow and lend securities
through approved intermediaries (SEBI, 2010). The SLBS is applicable in the cash market.
Table 4 shows the details of settlement of trade in the cash segment of NSE, such as the traded quantity,
deliverable quantity, and the percentage of deliverable quantity to traded quantity for the period 2001
2002 to 20102011. This table reveals that the total traded quantity of shares is high during the stock
market boom period (20032007). However, the delivered quantity of shares is much low compared to
the traded quantity. During 20002011, the average delivered shares was around 24% of total traded
quantity. This implies that the remaining 76% of trading is for speculative purposes for day trading. The
day trade can be either short-selling of securities or long-buying of securities, those who buy (long) will
square their long position or those who sell (short) will cover their short sales before market close.
Accordingly, for every short sale there will be a corresponding long (buy). Since there is limited
availability of data on short-selling in the Indian stock market, we use the above non-delivered quantity of
sharesthat is, 76%as a proxy for the size of short selling.3
3.2.2.2. Findings. In Table 5, we report the results of sectoral protability by simply assuming a short-selling
of 100%, which is close to what we estimate short-selling to be on the Indian stock market. The results are
reported for only those three strategies where an investor is able to take long and short positions by
forming an arbitrage portfolio with zero investment. The key ndings from the short-selling based trading
across three different strategies can be summarised as follows:
The IT sector is the most protable. The monthly average prots are in the 0.651.1% range.
The pharmaceutical sector is the least protable with prots/loss in the 0.37 to 0.02% range.
In terms of sectoral protability, all three strategies produce very consistent results. However, in terms
of risk, we notice a reduction when an investor moves from strategy II to a relatively more diversied
strategy. Consider the IT sector, for example. The standard deviation declines from 4.77% with strategy II
to 2.56% with strategy III.
Prots in the four protable sectors are much higher when short-selling is allowed compared to prots
obtained from strategies that do not allow for short-selling. Taking the evidence from strategy III, IT
prots are almost 172% higher under short-selling compared to a case of no short-selling. The second
largest gain in prots from short-selling is achieved by the energy sector (80%), followed by the FMCG
sector (40%). The smallest increase in prots with short-selling is experienced by the MNC sector whose
prots increase only by 4.5%.
When all sectors are considered as a portfolio, the portfolio returns from Strategies II to IV fall in the
[1.98%, 3.69%] range.
3.2.3. Results in the pre-crisis and crisis periods
Several studies (see, inter alia, Longstaff, 2010) have shown that the performance of stock markets has
been negatively affected by the 2007 global nancial crisis. This literature has analysed the stock markets
from different perspectives. We re-estimate sectoral prots by splitting the sample into a pre-crisis period
(30 August 2002 to 31 March 2007) and a period including the crisis (1 April 2007 to 31 December 2012).
3
While it is true that the magnitude of short-selling is unknown, the impact of short-selling on the Indian market has been of
interest. In a recent study, for instance, Giannikos and Gousgounis (2012) nd that during the period of short-sale ban (20012007)
the equity market was overpriced compared to the futures market. They argue that the main source of overpricing was opinion
dispersion. They further document that overpricing declined when the short-selling ban was lifted.

52

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

Table 4
Settlement in cash segment of NSE.
This table shows the details of settlement of trade in the cash segment of NSE, such as the traded quantity, deliverable quantity, and
the percentage of deliverable quantity to traded quantity for the period 20002001 to 20102011.
Source: SEBI Bullen (various months).
Year

Traded quantity
(lakh)

Deliverable quantity
(lakh)

% of delivered quantity to traded quantity

20012002
20022003
20032004
20042005
20052006
20062007
20072008
20082009
20092010
20102011
Average

274,695
365,403
704,537
787,996
818,438
850,510
148,123
141,893
220,588
181,091

59,299
82,353
175,546
202,276
227,239
239,070
367,970
275,270
47,481
49,737

21.59
16.5
24.92
25.67
27.76
28.11
24.84
21.59
21.86
27.47
23.6

The results for the pre-crisis period are reported in Table 6, while results from the crisis period are
reported in Table 7.
These results are interesting and reveal the following. First, in the pre-crisis period, there is mixed
evidence on protability. For example, while strategy III reveals protability for ve out of six sectors
(with the exception of the pharmaceutical sector), strategies II and IV reveal protability for only energy
and FMCG, and MNC and FMCG sectors, respectively. Compare this with the crisis period: strategies II, III
and IV consistently suggest that energy and IT are the only protable sectors. Second, the FMCG and IT
sectors remain the most protable in pre-crisis and crisis periods, respectively. Third, two sectors, namely,
MNC and FMCG, which were protable in the pre-crisis period become unprotable in the crisis period.

Table 5
Performance of long and short strategies with short-selling allowed (100%).
This table reports monthly mean returns based on short-selling. The row with Prob N 0 denotes the percentage to total months in
which prots from a given strategy exceeds zero. The row with Prob N market represents the percentage of total months in which
prots from a given strategy exceeds the market return. Here, we use the S&P CNX Nifty as a benchmark for the market. The paired
t-test indicates whether or not the return from each of the strategies is signicantly greater than the market return. ***, ** and *
represent statistical signicance at the 1%, 5% and 10% levels, respectively.
Banking

Energy

IT

MNC

FMCG

Pharmaceutical

Panel A: strategy II
Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N market (%)
Paired t-test

0.125
5.764
19.09
35.45
1.763

0.822**
4.167
14.55
35.45
0.824

0.889**
4.770
35.45
41.82
0.687

0.398*
2.194
4.55
34.55
1.692

0.433**
3.092
27.27
44.55
1.532

0.365
2.901
10.00
32.73
2.868

Panel B: strategy III

Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N market (%)
Paired t-test

0.823***
2.299
48.79
37.27
0.952

0.684***
2.061
34.55
33.64
1.208

1.138***
2.562
60.91
41.82
0.403

0.394***
1.296
38.18
30.91
1.779

0.630***
1.630
57.27
39.09
1.332

0.017
1.503
31.82
32.73
2.868

Panel C: strategy IV
Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N market (%)
Paired t-test

0.165
5.527
30.00
36.36
1.786

0.686**
3.710
20.00
34.55
1.045

0.645*
4.144
45.45
40.00
1.69

0.326*
1.572
16.36
32.73
1.853

0.402**
2.793
41.82
38.18
1.594

0.240
2.355
17.27
30.91
2.74

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

53

Table 6
Performance of long and short strategies with short-selling allowed (100%): Pre-crisis period.
This table reports monthly mean returns from short-selling over the pre-crisis period. The row with Prob N 0 denotes the
percentage to total months in which prots from a given strategy exceeds zero. The row with Prob N market represents the
percentage of total months in which prots from a given strategy exceeds the market return. Here, we use the S&P CNX Nifty as a
benchmark for the market. The paired t-test indicates whether or not the return from each of the strategies is signicantly greater
than the market return. ***, ** and * denote statistical signicance at the 10%, 5%, and 1% levels, respectively.
Banking

Energy

IT

MNC

FMCG

Pharmaceutical

Panel A: strategy II
Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N market (%)
Paired t-test

0.125
5.286
19.149
27.660
1.871

0.970*
3.585
12.766
29.787
0.778

0.159
4.358
31.915
29.787
1.489

0.600
2.627
6.383
27.660
1.673

0.992*
3.691
31.915
38.298
1.151

0.017
1.785
8.511
23.404
2.242

Panel B: strategy III

Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N Market (%)
Paired t-test

0.644***
2.093
19.149
36.170
1.154

0.850***
1.692
12.766
23.404
1.259

1.146***
2.858
31.915
34.043
0.857

0.594***
1.557
6.383
23.404
1.763

1.321***
1.758
31.915
31.915
0.926

0.272
0.923
8.511
25.532
2.209

Panel C: strategy IV
Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N market (%)
Paired t-test

0.167
5.527
25.532
34.043
1.867

0.351
3.710
19.149
27.660
1.194

0.125
4.144
46.809
27.660
1.687

0.517**
1.572
19.149
25.532
1.869

0.939*
2.793
46.809
27.660
1.265

0.222
2.355
14.894
23.404
2.209

Table 7
Performance of long and short strategies with short-selling allowed (100%): Crisis period.
This table reports mean monthly returns from short-selling over the crisis period. The row with Prob N 0 denotes the percentage to
total months in which prots from a given strategy exceeds zero. The row with Prob N market represents the percentage of total
months in which prots from a given strategy exceeds the market return. Here, we use the S&P CNX Nifty as a benchmark for the
market. The paired t-test indicates whether or not the return from each of the strategies is signicantly greater than the market
return. ***, ** and * represent statistical signicance at the 1%, 5% and 10% levels, respectively.
Banking
Panel A: strategy II
Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N market (%)
Paired t-test

0.017
6.110
19.048
41.270
0.776

Panel B: strategy III

Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N market (%)
Paired t-test

0.989***
2.431
50.794
38.095
0.049

Panel C: strategy IV
Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N market (%)
Paired t-test

0.121
5.986
34.921
37.089
0.034

Energy

IT

MNC

FMCG

Pharmaceutical

1.070*
4.533
15.873
39.683
0.389

1.327**
5.019
38.095
50.794
0.375

0.250
1.850
3.175
39.683
0.781

0.036
2.562
23.810
49.206
1.020

0.626
3.453
11.111
39.683
1.776

0.739**
2.290
36.508
41.270
0.502

1.145***
2.362
60.317
47.619
0.213

0.248
1.072
34.921
36.508
0.831

0.144
1.347
47.619
44.444
0.717

0.161
1.781
30.159
38.095
1.197

0.984***
4.454
46.032
49.206
0.465

0.192
1.420
14.286
38.095
0.157

0.035
2.357
38.095
46.032
0.176

0.547
2.677
19.048
36.508
0.291

0.901***
4.062
22.222
41.270
0.551

54

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

3.2.4. A robustness test

One aspect of our empirical analysis which remains unclear is that related to short-selling. The amount
of short-selling allowed on the Indian stock market is not made public. Therefore, in this paper, we
estimate it to be around 76%. This analysis on short-selling is, itself, a contribution to the literature. We
then allowed for short-selling of 100% in our empirical analysis, as reported in the previous section. We
want to test how prots will change if we reduce the amount of short-selling to something like 76%, which
is closest to what we believe is allowed on the Indian stock market. We do this and report the results in
Table 8.
The main ndings can be summarised as follows:
1. Like with 100% short-selling, four of the six sectoral prots are statistically signicant.
2. Like with 100% short-selling, the ranking of sectors does not change much, except that in two of the
three strategies, the energy sector becomes more protable than the IT sector.
3. Compared with 100% short-selling, there are only very marginal changes in the magnitude of prots for
some sectors, but not all. In three of the four protable sectors, prots are higher with 100%
short-selling compared to when short-selling is restricted to 76%.

3.3. Results from technical trading rule-based strategies

The results from technical trading prots based on the moving average, momentum and lter rules,
averaged across strategies and over time, are presented in Table 9. Two results are of particular relevance.
First, we notice that unlike the rank-based trading strategy prots, here returns from all six sectors are
protable; that is, they are statistically different from zero at the 1% level. This result holds regardless of
the month of holding period. Over the six-month holding period, for example, we notice that both
momentum prots (that is, winner minus loser prots) and the zero-cost portfolio prots (winner plus
loser prots) are maximised for the banking sector followed by the energy sector. Second, the magnitude
of prots is higher in all sectors compared to a rank-based momentum trading strategy as reported earlier.
Table 8
Performance of long and short strategies with limited short-selling allowed (maximum of 76% short-selling).
This table reports mean monthly returns from limited short-selling. The row with Prob N 0 denotes the percentage to total months
in which prots from a given strategy exceeds zero. The row with Prob N market represents the percentage of total months in
which prots from a given strategy exceeds the market return. Here, we use the S&P CNX Nifty as a benchmark for the market. The
paired t-test indicates whether or not the return from each of the strategies is signicantly greater than the market return. ***, ** and
* represent statistical signicance at the 1%, 5% and 10% levels, respectively.
Banking

Energy

IT

MNC

FMCG

Pharmaceutical

Panel A: strategy II
Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N Market (%)
Paired t-test

0.251
4.863
31.818
35.455
1.579

0.735**
3.680
17.273
37.273
0.970

0.852*
4.580
45.455
42.727
0.742

0.370*
2.048
13.636
34.545
1.729

0.579**
2.779
40.909
45.455
1.314

0.294
2.817
17.273
34.545
2.740

Panel B: strategy III

Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N Market (%)
Paired t-test

0.320
4.711
53.240
39.091
1.558

0.723**
3.379
56.364
33.636
1.008

0.705*
3.836
58.256
40.000
1.058

0.462**
2.032
60.000
32.165
1.569

0.586***
2.542
67.273
41.818
1.329

0.104
2.525
58.182
32.145
1.015

Panel C: strategy IV
Mean rets %
Std. dev.
Prob. N 0 (%)
Prob. N Market (%)
Paired t-test

0.443
4.926
54.545
38.182
1.297

0.727**
3.538
55.455
32.727
1.095

0.677*
3.901
59.091
40.909
1.064

0.524**
2.473
60.000
32.727
1.448

0.645**
2.671
66.364
40.000
1.226

0.294
2.728
58.182
31.818
1.805

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

55

Table 9
Average prots from technical trading rules.
This table reports average prots obtained from three different technical trading rules; namely, moving average (MA), momentum, and
lter rules. The MA rule is based on short-term MA and long-term MA of 1, 2, and 3 months, and 9 and 12 months, respectively. The
momentum rule generates a buy signal whenever current stock price is higher than or equal to m months ago, where we set m = 9 and
12, and a sell signal is generated whenever the current price is less than price m months ago. Finally, the lter rule generates a buy signal
whenever current return is greater than or equal to 110% of the average return based on tm, where we set m equals to 12.
Conversely, a sell signal is generated whenever current return is less than the 110% of average return based on tm. For individual stocks
in each of the six sectors, we identify winners and losers based on these buy and sell signals. We equally invest in winners in period t
and short in losers. Therefore, the momentum (mom) returns are simply returns of winners minus returns of losers and the portfolio
return is simply the returns of winners plus returns of losers. The corresponding standard deviation of momentum and portfolio returns
for each sector and for each holding periodthat is, for 1-month, 3-month, and 6-monthare also reported. ***, ** and * represent
statistical signicance at the 1%, 5% and 10% levels, respectively.
1-Month
Mom.

3-Month

6-Month

Portfolio

Mom.

Portfolio

Banking
Return
SD

2.341***
8.433

1.807***
7.405

2.491***
5.991

3.271*
5.003

2.569***
4.433

3.613***
3.529

Energy
Return
SD

1.773**
7.058

1.691***
6.283

1.979*
5.107

2.498***
4.423

2.087***
4.105

2.981***
3.31

FMCG
Return
SD

1.522***
5.114

1.574***
4.506

1.575***
3.471

2.124***
2.842

1.606***
2.698

2.290***
2.048

1.302*
7.77

0.863
7.53

2.404***
6.036

1.045*
5.848

3.398***
4.665

IT
Return
SD

0.334
11.26

MNC
Return
SD

1.303***
5.792

1.331***
5.112

1.409***
4.181

2.172***
3.42

1.498***
3.349

2.434***
2.638

Pharmaceutical
Return
SD

1.477***
5.537

1.075***
4.549

1.586***
3.863

1.897***
3.224

1.655***
2.884

2.010***
2.461

For each stock in a sector, winners and losers are identied from buy and sell signals generated using
the three technical trading rules explained earlier. As before, the buy and sell signals are generated every
month and prots are averaged (across the three technical trading strategies) every month, generating a
time-series of winner minus loser and zero-cost portfolios. This is to say that portfolios are re-balanced
every month. The momentum returns are computed as the difference between returns of winners and
losers whereas the portfolio return is computed as the sum of returns of winners and losers. The gures
reported in Panel A of Table 10 are these time-series averages and their standard deviation. All prots are
statistically different from zero at (at least) the 10% level of signicance.
Next we follow Jegadeesh and Titman (1993) and estimate prots based on buying winners and selling
losers. The idea is simple and proceeds as follows. Based on the geometric mean of returns sectors are
ranked from 1 to 6. Sectors ranked 1 and 2 are taken as winner sectors while the loser sectors are those
ranked 5 and 6. The strategy then is to invest equally in winner sectors and short equally in loser sectors,
holding the portfolios for 1-month, 3-month, and 6-month horizons. Portfolios are rebalanced every
month and momentum and zero-cost portfolio returns are computed. These results are reported in
panel B.
This is what we nd. First, with respect to individual stock momentum, we nd that: (a) average returns
from both winner minus losers and zero-cost portfolios offer positive and statistically signicant (at least at
the 5% level of signicance) returns in all sectors; and (b) prots, in terms of magnitude, increase with the
portfolio holding period. Second, when portfolios are held for three and six months, zero-cost portfolio
returns are maximised for the IT sector, followed by banking and MNC sectors. Moreover, at the 6-month

56

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

Table 10
Individual stock momentum by industry and industry momentum.
This table reports two sets of results organised into Panels A and B. In panel A, we have individual stock momentum return and
portfolio return and their standard deviations for 1-month, 3-month and 6-month holding periods. Using technical trading rules, buy
and sell signals are generated and winner and loser stocks are identied. Based on these winner and loser stocks, momentum (MP)
and zero-cost portfolios (ZCP) are generated. Panel B has the corresponding industry momentum and portfolio returns and their
standard deviations. Instead of ranking stocks, here we rank the six sectors. Sectors are ranked using the geometric average and
those ranked 1 and 2 are taken as winner sectors while the loser sectors are those ranked 5 and 6. The strategy then is investment
equally in winner sectors and shorting equally in loser sectors and holding the portfolios for 1-month, 3-month, and 6-month
horizons. Mom^ is risk-adjusted momentum returns following the proposal of Barroso and Santa-Clara (2013), which is simply a
target volatility (which is a constant and we set this to the sample standard deviation) multiplied by the actual momentum returns
over time normalised by a time-varying volatility, which we proxy using a rolling-window of six months. *** represent statistical
signicance at the 1%, 5% and 10% levels, respectively.
Panel A: individual stock momentum
1-Month

3-Month

6-Month

Banking

Return
SD

MP

Mom^

ZCP

3.392***
13.4150

3.7367***
11.1022

2.754***
11.3650

MP

Mom^

ZCP

MP
3.499***
8.7190

Mom^

ZCP

4.6709***
10.4182

4.381***
7.2420

Mom^

ZCP

MP
3.668***
6.5610

Mom^

Portfolio

6.7656***
10.0251

4.774***
5.5300

Mom^

ZCP

Energy

Return
SD

2.243***
11.8510

2.4928***
9.3465

1.684***
9.9900

MP
2.259***
6.8630

2.5613***
6.9064

2.872***
5.8580

MP
2.364***
5.2660

4.0199***
8.1776

3.280***
4.4170

FMCG
MP
Return
SD

2.166***
8.5630

Mom^
3.2973***
5.6463

ZCP
2.107***
7.6250

MP
2.209***
5.0160

Mom^
3.2844***
5.5595

ZCP
2.848***
4.2460

MP
2.302***
3.8420

Mom^
5.0557***
7.4907

ZCP
3.076***
3.1410

IT

Return
SD

MP

Mom^

ZCP

MP

Mom^

ZCP

MP

Mom^

ZCP

2.326***
15.2980

2.8890***
10.5348

2.668***
12.7600

2.968***
10.9770

4.1617***
11.9279

4.397***
9.0150

3.784***
10.7660

5.0551***
18.1241

5.744***
9.3280

MP

Mom^

ZCP

MP

Mom^

ZCP

MP

Mom^

ZCP

5.5097***
11.2295

3.523***
6.1630

Mom^

ZCP

MNC

Return
SD

2.472***
11.2170

3.9820***
8.9614

2.328***
9.5780

2.706***
7.3350

3.020***
6.2610

6.4161***
9.4742

4.073***
5.3640

Pharmaceutical
MP
Return
SD

2.462***
10.7040

Mom^
2.9173***
6.8730

ZCP
2.094***
9.1650

MP
2.596***
6.7720

3.6675***
6.5431

3.183***
5.8570

MP
2.842***
5.5040

Mom^
6.2759***
7.8865

ZCP
3.662***
4.8240

Panel B: industry momentum

1-month
MP
Return
SD

2.972***
12.3240

3-month
ZCP
2.568***
5.4860

MP
5.723***
11.4670

6-month
ZCP
0.4100
2.3580

MP
1.615***
0.4490

ZCP
2.055***
0.4490

holding period average returns from both winner minus loser and zero-cost portfolios are again maximised
for the IT sector, followed by the banking sector. Third, at the industry-level, average returns of all winners
minus loser portfolios are positive and statistically signicant at the 1% level. By comparison, the zero-cost
portfolio only offers positive and signicant monthly average returns in 1-month and 6-month holding
periods.

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

57

240
200
160
120

%
80
40
0
-40
2003 2004

2005

2006

2007

2008

2009

2010

2011

2012

Year
12 M
3M

1M
6M

24 M

Fig. 1. A time-series plot of average cumulative momentum prots. This gure plots the cumulative average momentum prots.
These prots are average over time (December 2003 to December 2012) for the six sectors for 1-month, 3-month, 6-month,
2-month, and 24-month horizons.

In addition, just to compare the performance over different holding periods, we plot in Fig. 1, average
(across sectors) cumulative momentum prots for 1-month, 3-month, 6-month, 12-month, and 24-month
horizons. We observe a clear pattern; that is, momentum prots for holding periods 1-month to 12-month
are fairly consistent but prots at the 24-month horizon are substantially different, and suggest a weaker
momentum.
On the whole, there are two clear messages from these individual and industry-based portfolio returns.
The rst point is about whether or not an investor should devise a trading strategy based on stocks or
sectors. There is mixed evidence on this. Consider the 1-month holding period rst. At the industry-level,
the momentum portfolio and the zero-cost portfolio offer investors an average return of 2.97% and 2.57%,
respectively, which are in excess of the corresponding sectoral average returns, except in the case of the
banking sector. Therefore, at a short holding period of 1-month it seems economically signicant for an
investor to trade on winner and loser sectors/industry as opposed to winner and loser stocks unless of
course the investor is only keen in investing in the banking sector. At the 3-month horizon, by
comparison, the momentum portfolio of the industry beats the average returns from winner minus loser
stocks in each sector, but the average industry zero-cost portfolio makes a loss of 0.41% per month
whereas all zero-cost sectoral portfolios make a statistically signicant prot. The second point relates to
the longest holding periodthe 6-month portfolio; here, both winner minus loser and zero-cost
portfolios of the industry offer less prots than corresponding prots from sectoral portfolios. Therefore,
where short horizon portfolios favour sectors over the industry as a whole, long horizon portfolios offer
the complete opposite implication for investors.
Following Barroso and Santa-Clara (2013), we also compute risk-adjusted momentum returns (Mom*),
which is simply a target volatility (which we take as the standard deviation over the analysis period)
multiplied by the actual momentum returns over time normalised by a time-varying volatility, which we
proxy using a rolling-window of six months. We generally observe that prots from both momentum and
zero-cost portfolios are less than the risk-adjusted prots; however, the results are robust in the sense that
average returns from both portfolios are positive and statistically different from zero.
In Table 11, we return to our average prots obtained from the three technical trading rules. All stocks
are collected and sorted on the basis of size (market capitalisation), book-to-market, and liquidity
(number of transactions). We do this in order to test the robustness of our industry-based results reported
in panel B of Table 10, motivated by empirical results presented in Moskowitz and Grinblatt (1999),
amongst others. Under each of these three control measures, quartiles of stocks are created. For example,
for size-based stocks, the rst quantile includes the largest 25% of stocks while quantile four includes the
smallest 25% of stocks in our sample. When performance is controlled by book-to-market ratio, the rst

58

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

Table 11
Prots sorted on size, book-to-market, and liquidity.
This table reports prots that are controlled for size (Panel A), book-to-market (Panel B) and liquidity (Panel C). Using combinations
of the three technical trading rules, the performance of portfolios were categorised by size (market capitalisation), book-to-market,
and liquidity (number of transactions). Quartiles are formed under each of the three control variables. For example, for size, the rst
quantile includes the largest 25% of stocks while quantile four includes the smallest 25% of stocks in our sample. When performance
is controlled by book-to-market ratio, the rst quartile includes the 25% of stocks which have the highest book-to-market ratio while
the fourth quantile includes the 25% of stocks with the lowest book-to-market ratio. Finally, when performance is judged on the
basis of liquidity, quartile one includes the 25% of most liquid stocks while the 25% of stocks with the least liquidity (that is, the 25%
most illiquid stocks) are in quantile four. For each quantile under each control measure, we identify winners and losers based on buy
and sell signals identied using technical trading rules. We equally invest in winners in period t and short in losers. From this
investment strategy, a time-series of momentum returns are computed (returns of winners minus returns of losers) and a
time-series of portfolio return is also computed (returns of winners plus returns of losers). This table reports the average of these
returns. The corresponding standard deviation of momentum and portfolio returns for each quartile for each holding periodthat is,
for 1-month, 3-month, and 6-monthare also reported. *** and ** represent statistical signicance at the 1%, 5% and 10% levels,
respectively.
1-Month
Panel A: size-based
Quantile 1
Return
SD
Quantile 2
Return
SD
Quantile 3
Return
SD
Quantile 4
Return
SD

3-Month

6-Month

Momentum
2.241***
10.299
Momentum
2.667***
11.781
Momentum
3.169***
12.410
Momentum
2.616***
13.582

Portfolio
1.896***
8.871
Portfolio
2.240***
9.885
Portfolio
2.882***
10.844
Portfolio
2.550***
11.525

Momentum
2.276***
6.091
Momentum
2.858***
6.949
Momentum
3.471
8.569
Momentum
2.917***
9.078

Portfolio
2.940***
5.133
Portfolio
3.526***
5.889
Portfolio
4.193
7.540
Portfolio
3.941***
7.561

Momentum
2.355***
4.534
Momentum
3.102**
1.803
Momentum
3.898***
7.409
Momentum
3.399***
8.373

Portfolio
3.228***
3.726
Portfolio
4.022***
5.820
Portfolio
4.884***
6.574
Portfolio
4.795***
7.174

Panel B: book-to-market-based
Quantile 1
Momentum
Return
1.885***
SD
10.378
Quantile 2
Momentum
Return
2.922***
SD
11.856
Quantile 3
Momentum
Return
2.655***
SD
11.867
Quantile 4
Momentum
Return
3.015
SD
13.802

Portfolio
1.927***
8.691
Portfolio
2.634***
10.211
Portfolio
2.271***
10.184
Portfolio
2.965
11.860

Momentum
2.051***
6.485
Momentum
3.094***
7.747
Momentum
2.885***
7.389
Momentum
3.303
8.893

Portfolio
2.944***
5.198
Portfolio
3.905***
6.572
Portfolio
3.551***
6.496
Portfolio
4.336
7.661

Momentum
2.258***
5.267
Momentum
3.354***
6.270
Momentum
3.172***
6.580
Momentum
3.776***
7.841

Portfolio
3.445***
4.286
Portfolio
4.387***
5.420
Portfolio
4.158***
5.694
Portfolio
5.054***
6.776

Panel C: liquidity-based
Quantile 1
Momentum
Return
2.651***
SD
11.640
Quantile 2
Momentum
Return
2.390***
SD
12.715
Quantile 3
Momentum
Return
2.780***
SD
12.205
Quantile 4
Momentum
Return
2.768***
SD
11.430

Portfolio
2.286***
9.955
Portfolio
2.125***
10.603
Portfolio
2.656***
10.492
Portfolio
2.641***
10.006

Momentum
2.786***
7.251
Momentum
2.514***
7.419
Momentum
3.059***
8.171
Momentum
3.052***
7.780

Portfolio
3.586***
6.131
Portfolio
3.416***
6.215
Portfolio
3.906***
6.979
Portfolio
3.850***
6.730

Momentum
2.978***
5.625
Momentum
2.733***
6.278
Momentum
3.436***
6.935
Momentum
3.513***
7.289

Portfolio
4.051***
4.689
Portfolio
3.950***
5.196
Portfolio
4.527***
5.997
Portfolio
4.564***
6.476

quantile includes the 25% of stocks which have the highest book-to-market ratio while the fourth quantile
includes the 25% of stocks with the lowest book-to-market ratio.
Finally, when performance is judged on the basis of liquidity, quantile one includes the 25% most
liquid stocks while the 25% of stocks with the least liquidity (that is, the 25% most illiquid stocks) are in

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

59

quantile four. For each quantile under each control measure, we identify winners and losers based on buy
and sell signals, identied using technical trading rules. We equally invest in winners in period t and
short in losers. From this investment strategy, a time-series of momentum returns is computed (returns
of winners minus returns of losers) and a time-series of portfolio returns is also computed (returns of
winners plus returns of losers).
There are two main ndings here. First, prots are positive and statistically signicant regardless of
how stocks are sorted. Therefore, that the Indian stock market is protable holds across a range of widely
used control measures. Second, the most highly liquid stocks (25% of the most liquid stocks) on the Indian
market offer the highest average returns from both the winner minus loser portfolio and the zero-cost
portfolio. This is true regardless of the holding period.
The nal part of our analysis is with regard to the difference in protability results which we obtain. Two
aspects of the Indian stock market have become clear. First, the market is protable. A range of different
strategies tend to offer investors statistically signicant monthly returns. Second, when we estimate prots by
sector, we observe that while most strategies support sectoral protability to hold for all six sectors, prots
tend to be sector-specic. This heterogeneity in sectoral prots is not surprising but is rather expected. This
type of sector-specic results can best be explained by the information diffusion hypothesis, proposed by
Hong et al. (2007), and empirically tested by Westerlund and Narayan (2014a), who show that different types
of nancial ratios predict returns of some sectors more successfully than others. Moreover, Hong et al. (2007)
test whether industry portfolios predict movements in stock markets. They nd that 14 out of 34 industries
can predict movements in stock markets, suggesting support for the information diffusion hypothesis. More
specially, they claim, on the evidence of this predictability, that the stock market reacts with a delay to
information contained in industry returns. Motivated by these ndings, we test whether market returns
predict sectoral returns on the Indian stock market. Essentially, our time-series predictive regression model
has the following form:
SRt MRt1 t
where SR is the sectoral stock return and MR is the market return lagged one period over time t which is
monthly from January 2001 to December 2012. The returns are computed as the natural log difference of
market price and stock price indices. The null hypothesis is that the market return does not predict
sectoral stock returns. To test for the null hypothesis, we estimate the model using the generalised least
squares estimator proposed by Westerlund and Narayan (2012, 2014b). The two main features relevant
for our predictive regression model are: (a) predictor endogeneity; and (b) heteroskedasticity. Both these
issues characterise predictive regression models where high frequency data are used and where returns
appear on both sides of the regression model. The results, reported in Table 12, are as follows. First, our

Table 12
Results on sectoral return predictability.
This table reports three ndings. Column 2 contains the FGLS test examining the null hypothesis that market return predicts sectoral
returns for India. Six sectors, namely, banking, energy, FMCG, IT, MNC, and pharmaceutical, are considered. The coefcient on beta
and the FGLS t-test statistic of Westerlund and Narayan (2012, 2014) are reported in parenthesis. Column 3 is about whether market
return for each of the six sectors is endogenous in our proposed predictive regression model. To test for endogeneity, the residuals
from the proposed predictive regression model are regressed on residuals obtained from the rst-order autoregressive model of
market returns. The slope coefcient is reported together in parenthesis with the t-test statistic examining the null hypothesis that
the slope coefcient is zero. The nal column reports the null hypothesis of no ARCH effects in market returns and in sectoral returns.
*** denotes statistical signicance at the 1% level.
Predictability test: = 0

Endogeneity test

Heteroskedasticity test (p-value)

Market returns

Sectoral returns

0.486***
0.281***
0.126**
0.869***
0.269***
0.264***

1.166***
0.920***
0.514***
1.714***
0.779***
0.669***

0.000
0.000
0.000
0.000
0.000
0.000

0.000
0.000
0.000
0.000
0.000
0.000

Sectors
Banking
Energy
FMCG
IT
MNC
Pharmaceutical

(4.894)
(3.186)
(1.941)
(3.323)
(3.772)
(3.880)

(20.115)
(20.928)
(11.290)
(5.825)
(22.364)
(13.991)

60

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

predictor variablemarket returnsis endogenous. This endogeneity is present in all sectoral predictive
regression models. Second, the ARCH-LM test on ltered returns based on an autoregressive model with
12 lags, suggests that the null hypothesis of no heteroskedasticity can be strongly rejected at the 1% level
for both the market returns and all the sectoral returns. Third, the null hypothesis of no predictability is
rejected strongly at the 1% level for all six sectors and the coefcient on predictability is in the range of
0.264 (pharmaceutical sector) and 0.869 (IT sector). This nding suggests that the sectoral returns react to
market returns with different speeds. For example, a 1% increase in market returns predicts a 0.869%
increase in the IT sector's returns but the effect of the market is almost four times smaller on the returns of
the pharmaceutical sector. This is suggestive of a gradual diffusion of information from the market to its
sectors.
4. Concluding remarks
In this paper we examine whether, using a range of trading strategies, investors can make prots from
the Indian stock market. Our contributions are three-fold. First, unlike the extant literature, we examine
protability at the sector-level; yet, we do not ignore the market. We employ a range of trading strategies,
allowing an investor to take long and short positions, and consistently nd that the IT sector is the most
protable. Second, a controversial issue about which not much public information is available relates to
the magnitude of short-selling. We provide an analysis and estimate the magnitude of short-selling on the
Indian market.
Third, we undertake a range of robustness tests, including identifying winners and losers from technical
trading rules and past performance, controlling for industry size, book-to-market, and liquidity factors, and
generating risk-adjusted momentum and zero-cost portfolio prots. All results point to the protability of the
Indian stock market. Importantly, we discover that some sectors are relatively more protable than others.
We show that this sector-specic protability is due to information diffusionthat is, the market returns,
while they do predict all six sectoral returns, the magnitude of predictability varies from sector-to-sector. So,
in our analysis, the source of information diffusion is not only the evidence that market returns predict stock
returns but also that the speed of predictability is different for different sectors.
There are two main implications emerging from our study. Both relate to prospects for future studies. First,
we believe that future studies, not only on the Indian stock market but on stock markets in general, should
show respect to sectoral heterogeneity in any hypothesis testing that involves sectors of a market. Second, in
light of lack of concrete information on short-selling in the Indian stock exchange, our estimate of the degree
of short-selling is merely a prediction. We emphasise on this point strongly, for we hope future studies will
improve upon our prediction error (if any), and, in doing so, our study sets the motivation for future research
on this subject. We are most excited about these future research outcomes.
References
Barroso, P., and Santa-Clara, P., (2013) Momentum has its moments, Unpublished Manuscript.
Chang, E.J., Lima, E.J.A., Tabak, B.M., 2004. Testing for predictability in emerging equity markets. Emerg. Mark. Rev. 5, 295316.
Dicle, M.F., Beyhan, A., Yao, L.J., 2010. Market efciency and international diversication: evidence from India. Int. Rev. Econ. Financ.
19, 313339.
Giannikos, C.I., Gousgounis, E., 2012. Short sale constraints and dispersion of opinion: evidence from the Indian equity market.
Financ. Rev. 47, 115143.
Gunasekarage, A., Power, D.M., 2001. The protability of moving average trading rules in South Asian stock markets. Emerg. Mark.
Rev. 2, 1733.
Hong, H., Torous, W., Valkanov, R., 2007. Do industries lead stock market? J. Financ. Econ. 83, 367396.
Jegadeesh, N., Titman, S., 1993. Returns to buying winners and selling losers: implications for stock market efciency. J. Financ. 48,
6591.
Kumar, K., Mamidi, V., Marisetty, V., 2011. Global markets exposure and price efciency: an empirical analysis of order ow
dynamics of NYSE-listed Indian rms. J. Int. Financ. Mark. Inst. Money 21, 686706.
Lee, C.I., Mathur, I., 1996a. Trading rule prots in European currency spot cross-rates. J. Bank. Financ. 20, 949962.
Lee, C.I., Mathur, I., 1996b. A comprehensive look at the efcacy of technical trading rules applied to cross-rates. Eur. J. Financ. 2,
389411.
Longstaff, F.A., 2010. The subprime credit crisis and contagion in nancial markets. J. Financ. Econ. 97, 436450.
Majumder, D., 2013. Towards an efcient stock market: empirical evidence from the Indian market. J. Policy Model 35, 572587.
Mishra, K.R., Sehgal, S., Bhanumurthy, N.R., 2011. A search for long-range dependence and chaotic structure in Indian stock market.
Rev. Financ. Econ. 20, 96104.
Moskowitz, T.J., Grinblatt, M., 1999. Do industries explain momentum? J. Financ. 54, 12491290.

P.K. Narayan et al. / Pacic-Basin Finance Journal 30 (2014) 4461

61

Narayan, P.K., Ahmed, H.A., 2014. Importance of skewness in decision making: evidence from the Indian stock exchange. Glob.
Financ. J. (in press).
Narayan, P.K., Narayan, S., Sharma, S., 2013. An analysis of commodity markets: what gain for investors? J. Bank. Financ. 37,
38783889.
Narayan, P.K., Narayan, S., P., K.P., 2014a. Stock returns, mutual fund ows and spillover shocks. Pac. Basin Financ. J. 29, 146162.
Narayan, P.K., Ahmed, H.A., Narayan, S., 2014b. Do momentum-based trading strategies work in the commodity futures markets? J.
Futur. Mark. http://dx.doi.org/10.1002/fut.21685.
Neely, C.J., Rapach, D.E., Tu, J., Zhou, G., 2013. Forecasting the equity risk premium: the role of technical indicators. Federal Reserve
Bank of St. Loius Working Paper Series (Working Paper 2010-008H).
NSE, 2011. Indian securities market: a reviewChapter 1 Securities Market in India An Overview, pp. 57.
Okunev, J., White, D., 2003. Do momentum-based strategies still work in foreign currency markets? J. Financ. Quant. Anal. 38,
425447.
Pathak, B., 2008. The Indian Financial System Markets, Institutions, and Services. Pearson Publication, pp. 215217 (incomplete).
SEBI, 1996. Report of the B.D. Shah Committee on Short-Sales. Securities and Exchange Board of India, Mumbai.
SEBI, 2010. Master Circular for Stock Exchanges on Trading. Securities and Exchange Board of India, Mumbai.
Szakmary, A.C., Mathur, I., 1997. Central bank intervention and trading rule prots in foreign exchange markets. J. Int. Money Financ.
16, 513535.
Westerlund, J., Narayan, P.K., 2012. Does the choice of estimator matter when forecasting returns? J. Bank. Financ. 36 (9), 26322640.
Westerlund, J., Narayan, P., 2014a. Testing for predictability in conditionally heteroskedastic stock returns. J. Financ. Econ. http://dx.
doi.org/10.1093/jjnec/nbu001.
Westerlund, J., Narayan, P., 2014b. A random coefcient approach to the predictability of stock returns in panels. J. Financ. Econ.
http://dx.doi.org/10.1093/jjnec/nbu003.