Sunteți pe pagina 1din 30

The current issue and full text archive of this journal is available on Emerald Insight at:

www.emeraldinsight.com/0972-7981.htm

JAMR
14,2 Evaluating alternative
performance benchmarks for
Indian mutual fund industry
222 Sanjay Sehgal
Department of Financial Studies, University of Delhi, Delhi, India, and
Sonal Babbar
Department of Commerce, Maitreyi College, University of Delhi, Delhi, India

Abstract
Purpose – The purpose of this paper is to perform a relative assessment of performance benchmarks based
on alternative asset pricing models to evaluate performance of mutual funds and suggest the best approach in
Indian context.
Design/methodology/approach – Sample of 237 open-ended Indian equity (growth) schemes from April
2003 to March 2013 is used. Both unconditional and conditional versions of eight performance models are
employed, namely, Jensen (1968) measure, three-moment asset pricing model, four-moment asset pricing
model, Fama and French (1993) three-factor model, Carhart (1997) four-factor model, Elton et al. (1999)
five-index model, Fama and French (2015) five-factor model and firm quality five-factor model.
Findings – Conditional version of Carhart (1997) model is found to be the most appropriate performance
benchmark in the Indian context. Success of conditional models over unconditional models highlights that
fund managers dynamically manage their portfolios.
Practical implications – A significant α generated over and above the return estimated using Carhart’s
(1997) model reflects true stock-picking skills of fund managers and it is, therefore, worth paying an active
management fee. Stock exchanges and credit rating agencies in India should construct indices incorporating
size, value and momentum factors to be used for purpose of benchmarking.
Originality/value – The study adds new evidence as to applicability of established asset pricing models as
performance benchmarks in emerging market India. It examines role of higher order moments in explaining
mutual fund returns which is an under researched area.
Keywords Mutual funds, Asset pricing, Conditional measures, Higher order moments,
Performance benchmarks
Paper type Research paper

1. Introduction
Mutual funds, in today’s volatile market environment, serve as key investment avenues for
investors by providing them with a safe and transparent platform to apportion their
investible funds in various securities and markets. The mutual fund industry in India is
more than five decades old. It has its genesis in 1964 with the establishment of the Unit
Trust of India by the Government of India. As the mutual fund industry gained momentum,
the Indian market witnessed the establishment of many public-sector funds from 1987
onwards and private-sector funds from 1993. In March 2016, 42 mutual fund houses were
operating in India (33 private, nine public) offering 2,420 schemes with assets under
management as at March 31, 2016 of 189.66 billion US dollars.
With the growth of the mutual fund industry in India, the performance evaluation of
mutual funds has become imperative. Every investor investing in mutual funds faces
questions such as, “Is this fund performing?” “Is a fund undeniably superior to its
counterparts?” “Does the fund outperform a passive buy-and-hold investment strategy?”
Journal of Advances in
Management Research The essence of performance evaluation lies in seeking answers to these questions and, in the
Vol. 14 No. 2, 2017
pp. 222-250
process, assessing performance benchmarks continuously.
© Emerald Publishing Limited
0972-7981
DOI 10.1108/JAMR-04-2016-0028 JEL Classification — G12, G23, C12, C32, C52
Literature on performance evaluation concentrates mainly on two skills of fund Indian mutual
managers – stock selection and market timing. The present study focuses on stock selection fund industry
skills, which involve the micro-forecasting of movements in prices of individual stocks and
identifying over or undervalued stocks relative to other equities in the market.
To capture the stock selection ability of mutual funds, Jensen (1968) developed a one-factor
model based on a capital asset pricing model (CAPM), which considered the market as the
only risk factor. The first extension of the CAPM takes place with the introduction of higher 223
order moments-based models which are employed as performance benchmarks for mutual
funds’ evaluation. Levy (1969), Jean (1971), Rubinstein (1973) and Scott and Horvath (1980)
have documented the significance of skewness and kurtosis of returns, along with the mean
and variance of returns when maximizing an investor’s expected utility. Several authors have
asserted that risk-averse investors prefer positive skewness to negative skewness in an asset
(Arditti, 1967; Tsiang, 1972; Kraus and Litzenberger, 1976; Brockett and Kahane, 1992;
Harvey and Siddique, 2000). Risk-averse investors, also, dislike assets that exhibit high
kurtosis and perceive that to be a negative investment incentive (Fang and Lai, 1997;
Hung et al., 2004; Lin and Wang, 2004; Javid, 2009, etc.). It has been documented that skewness
and kurtosis cannot be diversified away by increasing the sizes of portfolios (Arditti, 1971).
Authors have acknowledged that systematic skewness and systematic kurtosis, not total
skewness and total kurtosis are priced in markets. Kraus and Litzenberger (1976) developed a
three-moment CAPM by incorporating systematic skewness as an asset pricing factor for US
stocks. They noted that the utility function of a risk-averse investor is characterized by the
decreasing marginal utility of wealth and the non-increasing absolute risk aversion which
implies an aversion to variance and a preference for a positive return skewness in portfolios.
The empirical implication of their model is that the market price of systematic skewness will
have an inverse relationship with market skewness. Friend and Westerfield (1980) also
acknowledged that investors would be averse to positive co-skewness with the market, if the
market has negative skewness. Fang and Lai (1997), in addition, examined the effect of
co-kurtosis on asset pricing using a four-moment CAPM and documented that expected
returns were significantly related not only to covariance but also to co-skewness and -kurtosis.
Harvey and Siddique (2000) suggested two direct measures and two measures based on hedge
portfolios to compute co-skewness and found that co-skewness was important for explaining
equity returns. Moreno and Rodriguez (2005, 2009), and Ding and Shawky (2007) incorporated
co-skewness factors using measures suggested by Harvey and Siddique (2000) and found
that co-skewness played a significant role in explaining fund returns. Additionally,
Hung et al. (2004), and Hasan et al. (2013), etc. worked with a cubic market model consistent
with a four-moment model whereby they incorporated a co-kurtosis measure by regressing
excess stock returns on a cube of market risk premium and reported mixed results for
different markets. Recently, Gardijan and Skrinjaric (2015) have documented that higher order
moments play an important role in explaining variations in Croatian investment funds.
However, there is a second school of thought that concentrates on investment-style-based
multifactor models. Empirical finance research has documented that some factors, such as size
and book-to-market ratios can explain cross-sectional variations in asset returns. Particularly,
Fama and French (1993) (FF3F) have used size and book-to-market equity (BE/ME) factors to
develop a three-factor model for explaining asset returns. They mentioned the applicability of
their model for evaluating the performance of actively managed portfolios where the
interception of such a time series regression saw an average abnormal return after accounting
for three risk factors. Their three-factor model was capable of capturing much of the variation in
security returns not explained by CAPM. However, Jegadeesh and Titman’s (1993) short-term
momentum in stock returns remained unexplained by the model (Fama and French, 1996).
Carhart (1997) introduced stock momentum[1] as an additional factor as part of their devised
four-factor model. It explains a significant amount of variation in excess portfolio returns
JAMR relative to the CAPM and FF3F models. Elton et al. (1999) added a bond index stating that
14,2 many actively managed funds are invested in asset categories like long-term bonds, which are
not captured by stock market indices. Walkshausl (2013) associated the quality of firms as
factors of volatility effects. The FF3F model, augmented by a firm quality factor, has
contributed to the explanation of the volatility effect. Fama and French (2015) proposed a
five-factor model involving market, size, value, investment and profitability factors. According
224 to them, these new factors were derived naturally from Miller and Modigliani’s (1961) dividend
discount valuation model.
The aforementioned asset pricing models are unconditional performance measures based
on an assumption that expected returns and βs of funds remain constant over time and do
not account for changes in the state of an economy. Unconditional measures may
acknowledge the common time variation in a fund’s βs and expected market returns and act
as a superior performance indicator generated by the fund (Ferson and Schadt, 1996).
Acknowledging the dynamic nature of βs, Jagannathan and Wang (1996) examined
conditional versions of CAPM by incorporating three βs – for market, human capital and time
variability – and concluded that conditional CAPM explains the cross-section of returns much
better than an unconditional model. Hence, a third school of thought has its focus on the
conditional performance evaluation of mutual funds. Ferson and Schadt (1996) Ferson and
Schadt (1996) incorporated time-varying βs in existing performance measurement models to
generate αs conditioned on public information variables (IVs) – consistent with a semi-strong
form of market efficiency as interpreted by Fama (1970). Lee (1999), Sawicki and Ong (2000),
Fletcher (2002), Agudo et al. (2006) and Cortez et al. (2009), etc. found an improved,
risk-adjusted performance of funds by using a conditional approach.
Though many performance measurement models exist in the extant literature, there is no
consensus with respect to the optimal model. Studies like those of Otten and Bams (2004),
Fletcher and Kihanda (2005) and Lai and Lau (2010), etc. have focused on this aspect and
analyzed the effectiveness of alternative performance models amongst mutual funds.
In the Indian context, the literature has mainly focused on unconditional versions of
performance measures (Barua and Varma, 1991; Jayadev, 1996; Chandel and Verma, 2005;
Sehgal and Jhanwar, 2008; Shanmugham and Zabiulla, 2011, etc.). A handful of studies have
also concentrated on conditional performance evaluation (Roy and Deb, 2004; Dhar, 2013;
Roy, 2015a, b, 2016). However, no study has comprehensively evaluated alternative asset
pricing models for mutual fund performance evaluation in India. The present study has
attempted to fill this important research gap by covering a wide range of performance
benchmarks based on alternative asset pricing models using both unconditional and
conditional performance evaluation approaches.
Thus, the objective of the study is to find the optimal benchmark for mutual funds’
performance evaluation in the Indian context by evaluating performance measures based on
alternative asset pricing models.
The paper is organized as follows: Section 2 provides a review of relevant literature.
Section 3 discusses mutual fund performance models. Section 4 describes the data and their
sources. Section 5 discusses an estimation procedure. Section 6 assesses unconditional
performance measures of mutual funds. Conditional performance measures are discussed in
section 7. Section 8 presents a comparison between unconditional and conditional measures.
The last section provides a summary and concluding observations.

2. Review of literature
2.1 International studies
Treynor (1965) and Sharpe (1966) established Treynor and Sharpe ratios, respectively, to
rate the management of investment funds. Jensen (1968) developed an absolute measure
of performance to capture the stock selection skills of fund managers by utilizing CAPM.
An analysis of annual returns, net of expenses, for 115 open-ended mutual funds suggested Indian mutual
that, on average, fund managers exhibited poor stock selection abilities. Many studies fund industry
have been conducted by various authors in different time frames and market contexts
leading to mixed conclusions with respect to the stock selection capabilities of fund
managers (Ippolito, 1989; Grinblatt and Titman, 1989; Elton et al., 1993; Malkiel, 1995;
Swinkels and Rzezniczak, 2009; Neto, 2014; Naz et al., 2015).
Working in a mean-variance-skewness framework, Kraus and Litzenberger (1976) 225
extended CAPM to incorporate the effects of skewness on asset pricing. Their estimate of
the market price of systematic skewness was significant and negative, as predicted by the
three-moment CAPM. Friend and Westerfield (1980) affirmed that investors may pay a
premium for positive skewness in their portfolios. Sears and Wei (1985) considered an
elasticity coefficient relating risk to skewness where skewness preference was independent
of the effects of a market risk premium. Fang and Lai (1997) reported the significant impact
of systematic skewness and kurtosis on asset pricing using a four-moment CAPM.
Hwang and Satchell (1999) documented that emerging market returns are better explained
using co-skewness and co-kurtosis. Harvey and Siddique (2000), Lin and Wang (2003) and
Barone Adesi et al. (2004) documented the significance of co-skewness in asset pricing.
Christie and Chaudhry (2001) analyzed the contribution of co-skewness and co-kurtosis on
the returns from futures markets and co-moments increased the explanatory power of
return-generating processes in futures markets. Liow and Chan (2005) maintained that
co-kurtosis was significant for explaining returns from securities in real estate markets. Moreno
and Rodriguez (2005, 2009), Ranaldo and Favre (2005) and Ding and Shawky (2007) confirmed
the importance of co-skewness and -kurtosis for evaluating the performance of mutual funds
and hedge funds. These results have been confirmed by Javid (2009), Doan et al. (2010),
Gardijan and Skrinjaric (2015) and Ajibola et al. (2015) for different markets.
Nonetheless, major shifts in methodologies were observed in the 1990s with the advent of
multifactor models based on firm characteristics. The Fama and French (1993, 1996) three-
factor model, consistent with Merton’s (1973) intertemporal CAPM and Ross’s (1976)
arbitrage pricing theory, incorporated size and book-to-market equity factors to develop a
three-factor model to explain stock returns. Carhart (1997) incorporated momentum in stock
returns as an additional factor and the four-factor model explained a significant amount of
variation in excess portfolio returns. Subsequently, Elton et al. (1999) added a bond market
index. Recently, Fama and French (2015) have included investment and profitability
patterns in average stock returns. The five-factor model explains between 71 and 94 percent
of cross-sectional variations in expected returns from portfolios.
Jagannathan and Wang (1996) and Ferson and Schadt (1996) advocated conditional
performance evaluation as it captures the time-varying properties of βs and found
improvements in the performance of investment funds. Christopherson et al. (1998)
considered the dynamic character of αs as well as βs. Lee (1999), Sawicki and Ong (2000),
and Fletcher (2002) observed superior results on using a conditional approach
while Dahlquist et al. (2000), and Bessler et al. (2009) found that opposite results
pertained in their studies.
Testing the effectiveness of performance measures, Fletcher and Forbes (2002),
Otten and Bams (2004) and Lai and Lau (2010) concluded in favor of Carhart’s (1997) model
while Fletcher and Kihanda (2005) and Messis et al. (2007) selected a four-moment CAPM as
part of a higher order framework.

2.2 Indian studies


Barua and Varma (1991), Chander (2005) and Chandel and Verma (2005) reported
on the superior stock selection skills of fund managers. By contrast, Jayadev (1996),
and Anand and Murugaiah (2006) observed a lack of selectivity skills amongst them.
JAMR Sehgal and Jhanwar (2008) propagated the use of daily data while analyzing the
14,2 performance of mutual funds. According to them, previous studies that reported weak
results concerning stock selectivity could have been distorted by the use of a lower
frequency of observations, which may have been one of the major reasons for their
outcomes. Kaur (2013) reported positive selectivity skills, while Chaudhry and Shakeel
(2014) found poor skills when using daily data. Evaluating performance in a conditional
226 setting, Roy and Deb (2004) and Roy (2015, 2016) observed improved selectivity
performance using a conditional Jensen measure. Dhar (2013), using unconditional and
conditional Jensen measures, did not find much difference between the stock selection
abilities of fund managers.

2.3 Research gap


In the Indian context, studies have mainly focused on unconditional measures of performance,
particularly Jensen’s (1968) measure. Though, a few studies have directed their attention
toward conditional measures, they have focused on conditional versions of Jensen (1968) only
for measuring stock selectivity abilities. Notably, the application of more recently developed
multifactor models, such as Fama’s and French’s (2015) model, Walkshausl’s (2013) model
involving a firm quality factor (also adapted by Pandey and Sehgal, 2017) and higher order
moments-based models, such as the three-moment asset pricing model (applied by Ding and
Shawky, 2007), and the four-moment asset pricing model (applied by Lin and Wang, 2004;
Hung et al., 2004, etc.), for assessing mutual funds’ performances, is virtually absent from
Indian literature. Moreover, Indian studies have primarily used lower observation frequency
(monthly data) while analyzing performance. Besides, there is no study which collates
alternative asset pricing models and comprehensively examines their efficacy. The present
study attempts to fill this important research gap in mutual fund performance evaluation
literature for India by examining a wide range of performance benchmarks based on
alternative asset pricing models. It evaluates both unconditional and conditional versions of
these performance measures using the daily frequency of observations for achieving clear
performance evaluation results, as recommended by Goetzmann et al. (2000), Bollen and
Busse (2001) and Gallefoss et al. (2015). According to them, mutual fund managers tend to
rebalance their portfolios with higher not (lower frequency), thus, daily data can capture the
contributions made by managers’ skills in a better manner than is the case for monthly data.

3. Model specification
In this study, the performance evaluation for mutual funds is based on three types of
models, namely, a one-factor performance measure based on CAPM, higher order moments-
based models and investment-style, characteristics-based, multifactor models. These models
have been examined using unconditional and conditional approaches.

3.1 Unconditional measures of performance


3.1.1 Jensen (1968) measure ( JENSEN). One-factor CAPM, as proposed by Sharpe (1964) and
Lintner (1965) was utilized by Jensen (1968) to develop a measure of portfolio performance given
as follows:

Rpt Rf t ¼ ap þbp ðRM t Rf t Þ þept (1)

where Rpt − Rft ¼ excess returns on mutual fund portfolio p at time t, RMt − Rft ¼ excess
returns on market index at time t, αp ¼ intercept term measuring stock selection ability of
portfolio p, βp ¼ sensitivity coefficient representing covariance of portfolio p relative to market,
ept ¼ error term.
3.1.2 Three-moment asset pricing model (3M). Harvey and Siddique (2000) suggested an Indian mutual
approach for estimating co-skewness by regressing excess stock returns on the square of market fund industry
excess returns in a stochastic discount factor pricing framework. Ding and Shawky (2007)
utilized the Harvey and Siddique (2000) measure for evaluating hedge fund performance in the
following form (assuming that market return and squared market return were orthogonalized):
Rpt Rf t ¼ ap þbp ðRM t Rf t Þþgp ðRM t Rf t Þ2 þept (2)
227
where βp, γp ¼ sensitivity coefficients representing the co-variance and -skewness of portfolio p
relative to the market.
Friend and Westerfield (1980), Harvey and Siddique (2000) and Ding and Shawky (2007),
etc. affirmed that investors have a preference for positive skewness in their portfolios and,
as a result, they desire for either the negative or positive co-skewness of assets with market
portfolios depending on the skewness of market portfolios.
3.1.3 Four-moment asset pricing model (4M). Hung et al. (2004) worked with a cubic
market model consistent with a four-moment model whereby they incorporated co-kurtosis
measures by regressing excess stock returns on a cube of market risk premium (i.e. an
adapted Harvey and Siddique (2000) measure of co-skewness). Furthermore, many studies
have made use of a cubic market model to incorporate co-kurtosis and explain returns on
portfolios. For example, Lin and Wang (2004), Javid (2009), Hasan and Kamil (2014), Gardijan
and Skrinjaric (2015) and Ajibola et al. (2015), etc. reported mixed results for different markets.
The present study employed a four-moment asset pricing model with the following form:
Rpt Rf t ¼ ap þbp ðRM t Rf t Þ þgp ðRM t Rf t Þ2 þdp ðRM t Rf t Þ3 þ ept (3)
where βp, γp, δp ¼ sensitivity coefficients represented the covariance, co-skewness and co-
kurtosis of portfolio p relative to the market.
Investors have a negative preference for kurtosis. Hence, they would be averse to
positive co-kurtosis in a market if a market portfolio was characterized by leptokurtosis
(Gregoriou et al., 2003).
3.1.4 The Fama and French (1993) three-factor model (FF3F). Fama and French (1993)
incorporated size and the book-to-market equity ratio along with market factors. In the
following regression, the price-to-book ratio (P/B) was used in place of the book-to-market
ratio (BE/ME) and a low minus high (LMH P/B) factor was constructed in place of a high
minus low (HML BE/ME) factor:
Rpt Rf t ¼ ap þbp ðRM t Rf t Þþsp ðSMBt Þþl p ðLMHt Þ þ ept (4)

where SMBt (small minus big) and LMHt (low minus high) represented size and value
factors, respectively; βp, sp, lp ¼ sensitivity coefficients associated with market, size and
value factors, respectively.
Fama and French (1996) stated that firms with small MCAP and high BE/ME ratios have
performed poorly and are vulnerable to financial distress and, thus, such firms commanded
a risk premium.
3.1.5 The Carhart (1997) four-factor model (CARHART). Carhart (1997) added Jegadeesh
and Titman’s (1993) momentum factor besides market, size and value factors. The model
explained a significant amount of variation in excess portfolio returns relative to the CAPM and
FF3F models:
Rpt Rf t ¼ ap þbp ðRM t Rf t Þ þsp ðSMBt Þþl p ðLMHt Þþ wp ðWMLt Þþept (5)

where WMLt (winners minus losers) represented a momentum factor; βp, sp, lp, wp ¼ sensitivity
coefficients associated with market, size, value and momentum factors, respectively.
JAMR 3.1.6 The Elton et al. (1999) five-index model (ELTON). Elton et al. (1999) proposed the
14,2 inclusion of a long-term bond index in mutual fund performance assessments stating that
some actively managed funds invested in asset categories like long-term bonds, which were
not captured by the stock market index.

Rpt Rf t ¼ ap þbp ðRM t Rf t Þþsp ðSMBt Þþl p ðLMHt Þ


228 þ wp ðWMLt Þþbp ðRBt Rf t Þþ ept (6)
where RBt − Rft ¼ excess returns on the Government Securities Index at time t. βp, sp, lp, wp,
bp ¼ sensitivity coefficients associated with market, size, value, momentum factors and the
bond index, respectively.
3.1.7 The Fama and French (2015) five-factor model (FF5F). Fama and French added
investment and profitability factors to FF3F. For the investment factor, the conservative
minus aggressive (CMA) factor and for the profitability factor, the robust minus weak factor
was constructed by them. Firms with higher capital investment are likely to have negative
adjusted future returns, indicating that investors tend to react negatively to “empire-
building” attitudes (Titman et al., 2004).
More profitable firms are expected to outperform less profitable firms as profits are
considered to be rewards for growth and innovation, which exposes entrepreneurs to
greater risk, thus resulting in higher returns. However, the present study used weak minus
robust (WMR) as a profitability factor following Sehgal and Subramaniam (2012), who
found that low-profitability stocks outperformed high-profitability stocks on the Indian
stock market[2], which was contrary to Fama’s and French’s (2015) findings:

Rpt Rf t ¼ ap þbp ðRM t Rf t Þþsp ðSMBt Þþl p ðLMHt Þ

þ ip ðCMAt Þþpp ðWMRt Þþept (7)


where CMAt and WMRt represent investment and profitability factors, respectively. βp, sp, lp,
ip, pp ¼ sensitivity coefficients associated with market, size, value, investment and
profitability factors, respectively.
3.1.8 Firm quality five-factor model (FQUAL). Walkshausl (2013) associated the quality
of firms to volatility effects (high returns to low-volatility stocks). The FF3F model was
augmented by a firm quality factor using two proxies, namely, cash flow variability and
profitability. Pandey and Sehgal (2017) adopted the said model for the Indian stock market
and confirmed the role of the firm quality factor in explaining volatility patterns in stock
returns. They found cash flow variability to be a more appropriate measure of firm quality.
Hence, the present study has used the cash flow variability measure as a proxy for firm
quality and replaced the profitability factor in the FF5F model. Firm quality is measured as
a difference between the average returns on portfolios of low and high cash flow variability
stocks. Firms of high-quality generate significantly higher average future returns since they
tend to generate more stable cash flows and are less prone to distress:

Rpt Rf t ¼ ap þbp ðRM t Rf t Þþsp ðSMBt Þþl p ðLMHt Þ

þ ip ðCMAt Þþqp ðQCFVt Þþept (8)


where QCFVt (quality cash flow variability) represents the firm quality factor. βp, sp, lp, ip,
qp ¼ sensitivity coefficients associated with market, size, value, investment and firm quality
factors, respectively.
Remaining terms have the same meanings as in previous equations.
3.2 Conditional performance measures Indian mutual
The study uses the conditional approach of performance evaluation (CPE) as proposed in fund industry
Ferson and Schadt (1996). They proposed the following conditional form of model based on
CAPM incorporating a time-varying beta βpm(Zt − 1) of portfolio p which was conditional on
an observable set of instruments ( Zt − 1):

Rpt Rf t ¼ bpm ðZ t1 Þ ðRM t –Rf t Þþupt (9)


229

Eðup;t =Z t1 Þ ¼ 0 (10)

Eðup;t ; ðRM t –Rf t Þ=Z t1 Þ ¼ 0 (11)

i.e. errors on average are zero and not correlated with a market risk premium. Where,
Rpt − Rft ¼ excess returns on managed portfolio p at time t, RMt − Rft ¼ excess returns on
market index at time t, βpm(Zt − 1) ¼ conditional βs of portfolio p that depend on information
vector (Zt − 1) at time t − 1.
Primarily, this class of models assumes + that returns and risks in time t can be predicted
at time t − 1, utilizing publicly available information variables (IVs) whose values are realized
in time t − 1. The core idea of CPE is to eliminate from the performance measure an effect of
investment strategy which may be replicated using publicly available information, consistent
with the semi-strong form of market efficiency as interpreted by Fama (1970).
Time variations in βs are incorporated in performance models using IVs. The β of the
fund is assumed to be a linear function of IVs, namely, dividend yield (DY), treasury bills
yield (TB), the slope of term structure (TS) and a corporate default risk spread (DRS). All IVs
were demeaned and used in their lagged form.
3.2.1 Conditional version of the Jensen’s (1968) measure. The conditional version of the
Jensen measure, based on CAPM, was estimated in the following form:
 
Rpt Rf t ¼ ap;c þbp;mkt ðRM t –Rf t Þþbp;dy ðRM t –Rf t Þ n ðdyt1 Þ
   
þbp;tb ðRM t –Rf t Þ n ðtbt1 Þ þ bp;ts ðRM t –Rf t Þ n ðtst1 Þ
 
þbp;drs ðRM t –Rf t Þ n ðdrs:t1 Þ þ A pt (12)
where, αp, c ¼ intercept term, conditional α, which measures the stock selection ability of a
mutual fund portfolio p. It is the average difference between actual returns on a mutual fund
portfolio and the expected returns on a dynamic investment strategy, conditional on the
state of the economy. βp, mkt ¼ the average of the conditional β of portfolio p, βp, dy, βp, tb, βp, ts,
βp, drs ¼ coefficients of the product of excess market returns and each lagged demeaned IV to
capture the characteristics of an economy while evaluating performance, dyt − 1, tbt − 1, tst − 1,
drst − 1 ¼ demeaned IVs. ∈pt ¼ an error term.
The remaining variables have the same meanings as in Equation (1)
This new regression may be interpreted as a multiple regression model in an
unconditional setting wherein the first variable is a market factor and subsequent variables
are products of lagged IVs with the market factor, which accounts for the time-varying
property of a β.
Conditional versions of the remaining performance models were constructed analogous
to Equation (12). Concisely, time-varying βs were introduced using interaction terms of risk
factors mentioned in Equations (2) to (8) and the demeaned IVs.
JAMR 4. Data
14,2 4.1 Mutual funds data
The study focused on equity-based schemes. Close-ended, equity-based schemes were not
included owing to the fact that their market prices become more relevant than net asset
values (NAVs) in the intermediate period, and their former values are significantly affected
by market sentiments. Further, these prices need not converge to NAVs even around the
230 maturity phase of schemes, that are typically referred to as close-ended fund puzzles[3].
As at March 31, 2013, there were 292 open-ended equity schemes in India. We retained
equity schemes that had a minimum of three years of data over the ten-year study period,
i.e. April 2003-March 2013[4]. This resulted in a final sample of 237 open-ended equity
mutual fund schemes which had growth as their objective. The choice of period lay in the
fact that the majority of schemes, about 155 out of 237 (65 percent), were launched after
April 1, 2003, signifying this as a suitable beginning place for the sample period. The study
period experienced a structural break because of the global financial crisis in August 2007.
Therefore, the total sample period was divided into two subperiods. August 9, 2007, when
the global financial recession set in, was chosen as the demarcation date (Filardo et al., 2010).
For Subperiod I (April 1, 2003-August 8, 2007), the number of mutual fund schemes was 199
and for Subperiod II (August 9, 2007-March 31, 2013), it was 237, as 38 schemes were
launched in Subperiod II. Daily dividend adjusted NAVs for sample schemes were drawn
from MFI Explorer, the mutual fund database offered by ICRA Online Limited. Daily NAVs
are used to calculate percentage returns, which are employed in subsequent estimation
procedures. Sample observations varied from scheme to scheme as certain mutual fund
schemes were launched subsequent to April 2003. Thus, observations considered for such
schemes ran from their dates of inception until March 2013.

4.2 Benchmark indices


The S&P Bombay Stock Exchange 500 Index (henceforth, the BSE 500 Index) was used as a
proxy for the market index. It represents about 93 percent of the total market capitalization
on the BSE. It covers 20 major industries in the economy. The BSE 500 Index is constructed
on the lines of the S&P 500 in the USA. Daily index values were used to estimate percentage
daily index returns.
BSE 500 Index stocks were used to construct size, value, momentum, investment,
profitability and firm quality factors on a daily basis. Stock prices used were adjusted for
capitalization changes, such as stock dividends, stock splits and rights issues. Stock prices
were used to compute percentage daily returns on sample securities.
Market capitalization (MCAP is the total market value of a company’s outstanding
shares) of securities and price-to-book value (P/B ratio is the security’s market price divided
by its book value) were used for constructing size and value factors. The natural log of
the MCAP and P/B ratio at the end of March of year t were used. Daily closing adjusted
share prices of the BSE 500 index stocks for the period April of year t ‒ 1 to March of Year
t were used for constructing the momentum factor. Growth in total assets and profit after
tax divided by net worth (PAT/Networth), both taken at the end of March of year t, were
used to construct investment and profitability factors, respectively. Cash flow variability
was estimated as a standard deviation of total cash flows (cash flows from operating,
investing and financing activities) over the last five years prior to a portfolio’s formation.
While annual data for the MCAP and P/B ratio had been obtained from March 2003
onwards, annual data for total assets for investments was taken from March 2002
(to estimate annual growth) and cash flows were collected from March 1998 (to estimate
the five-year cash flow variability). Stock price data were collected from March 2002
(to estimate price momentum). The data source was CMIE Prowess, a financial software
popularly used in India.
Additionally, the Government Securities Index, as compiled by the National Stock Indian mutual
Exchange of India (NSE) was included to compute daily returns on the bond index. The data fund industry
source for this was the NSE website. The 91-day treasury-bills rate was used as a risk-free
proxy and compiled from the official website of the Reserve Bank of India.

4.3 Public IVs


Economic variables recommended by Ferson and Schadt (1996) were widely used in 231
several studies for predicting security returns and risks over time (Sawicki and
Ong, 2000; Roy and Deb, 2004; Otten and Bams, 2004; Moreno and Rodriguez, 2009, etc.).
Following them, the variables used here were as follows: DY was defined as the
previous 12 months’ dividend payments for index stocks divided by price level at the
end of the previous month. DY on the CNX 500 Index was collected from the NSE website;
the treasury bill yield (TB) was the yield on the Treasury Bill Index as compiled
by the NSE. The data source was the NSE website; the term structure of interest
rates (TS) was computed as the difference of yields on the ten-year government
bond index (long-term bonds) and the Treasury Bill Index (short-term bonds). Data on
the ten-year government bond index and the Treasury Bill Index were collected from
the Bloomberg database and the NSE website, respectively; (d) corporate DRS in the
corporate bond market was measured as the difference of yield on BBB-rated ten-year
corporate bonds (low-grade bonds) and AAA-rated ten-year corporate bonds (high-grade
bonds). Data on DRS were collected from the Bloomberg database and were available
from July 2004.

5. Estimation procedure
We estimated daily excess returns for each sample fund. Since NAV is computed after
allowing for scheme expenses, the percentage change in NAV would generate total returns
that were net of expenses and management fees but were gross of any load charges. Excess
returns on funds were estimated as the difference between NAV-based percentage returns
and risk-free rates of return.
Excess returns on the market index were similarly estimated. The square and cube of
excess returns on the market index were the risk premiums associated with the co-skewness
and co-kurtosis of portfolios relative to the market, respectively.
Excess returns on the government’s bond index were estimated as the differences
between yields on the long-term government securities index and the risk-free rates
of return.
The fundamental based factors were constructed as follows:
• Size and value factors.
Following Fama and French’s (1993) methodology, four portfolios were formed from the
intersection of two MCAP and two P/B groups[5] of BSE 500 Index stocks, namely,
SMALL-LOW (S/L), SMALL-HIGH (S/H), BIG-LOW (B/L) and BIG-HIGH (B/H), where
SMALL-LOW (S/L) represented small cap and low P/B stocks and so on. Daily returns
were calculated on each equally weighted double-sorted portfolio for the period from April
of year t to March of year t+1. The portfolios were updated annually to 2012 starting from
2003, and at the end of March of every year new portfolios were formed using fresh data on
MCAP and P/B for the BSE 500 securities.

Small minus big represented the size factor, where it was computed as:

SMB ¼ ðS=LþS=H Þ=2ðB=L þB=H Þ=2 (13)


JAMR This size factor was free from value effects by construction. Similarly, low minus
14,2 high (LMH) represented the value factor which was independent of the size factor as
given below:

LMH ¼ ðS=LþB=LÞ=2ðS=H þB=H Þ=2 (14)

232 • Momentum factor.


To construct Carhart’s (1997) momentum factor, sample stocks were ranked on the basis of
average daily returns computed from April of year t − 1 to March of year t, i.e. a year
immediately prior to a portfolio’s formation. These ranked securities were then classified
into five equally weighted portfolios – P1 to P5 – which were held for the next 12 months.
The top 20 percent of past performers was denoted as “P5, winners portfolio” and the
bottom 20 percent, based on past returns, were denoted as “P1, losers portfolio.” Portfolios
were rebalanced annually to 2012, starting from 2003. The momentum factor, winners
minus losers, was calculated as the difference between average returns on winners’ (P5) and
losers’ portfolios (P1):
• Investment, profitability and firm quality factors.
These factors were constructed similarly to those for the momentum factor[6]. For the
construction of investment, profitability and firm quality factors, BSE 500 Index stocks
were ranked on the basis of the growth in total assets, PAT/networth, and cash
flow variability, respectively, at the end of March of year t [7]. Following Pandey and
Sehgal (2017), cash flow variability at the end of March of year t was computed as the
standard deviation of cash flows from operations for five consecutive years prior to a
portfolio’s formation. The top 20 percent based on growth in total assets, profitability and
cash flow variability, were referred to as “P5” and named an aggressive investment
portfolio, a robust portfolio, and a high-variability portfolio, respectively, and the bottom
20 percent based on growth in total assets, profitability and cash flow variability form were
referred to as “P1” and named a conservative investment portfolio, a weak portfolio, and a
low variability portfolio, respectively. Daily returns were calculated on each equally
weighted portfolio for the period from April of year t to March of year t + 1. The portfolios
were rebalanced annually to 2012 starting from 2003, where at the end of March of every
year, new portfolios were formed using fresh data on growth in total assets, profitability and
cash flow variability for BSE 500 securities. The investment factor was constructed as
CMA; the profitability factor as WMR; and the firm quality factor as LMH cash flow
variability portfolios.
Multiple regression analysis was conducted using the ordinary least squares method for
unconditional and conditional models. In each model, an intercept dummy variable was
added to account for the global financial crisis. Before August 9, 2007, the dummy’s value
was “0” and after this date, it assumed a value of “1.” The statistical significance of
parameters was tested at the 5 percent level (two-tailed basis). To deal with multicollinearity
among explanatory variables, the pair wise correlation between them was computed and
corrections were made for multicollinearity[8] following the approach suggested by
Giliberto (1985) and Lance (1988). Wherever required, White (1980) heteroskedasticity
consistent standard errors and Newey and West (1987) heteroscedasticity and autocorrelation
consistent standard errors were estimated.
The adjusted R2), the Akaike information criterion (AIC), Schwarz’s Bayesian
criterion (SBC) and the log likelihood ratio test (Log L) were employed as model selection
criteria. In model selection, higher values of adjusted R2, and smaller values of AIC and SBC
are preferred.
The goal of model selection criteria is to select a model that best balances parsimony and Indian mutual
fit with the observed data. Indicators, namely, adjusted R2, AIC and SBC, impose a penalty fund industry
for including an increasingly large number of regressors. AIC imposes a harsher
penalty than adjusted R2 for adding more regressors (Gujarati, 2004). SBC imposes a
larger penalty than AIC for additional parameters. It may be noted that AIC and SBC share
the same goodness-of-fit term, but the penalty term of SBC is potentially much more
stringent than the penalty term of AIC. Thus, SBC tends to choose fitted models that are 233
more parsimonious than those favored by AIC (Kadane and Lazar, 2004; Cavanaugh, 2009).
On the other hand, the indicators AIC, SBC and Log L consider likelihood function in one
form or other, but the information criteria (AIC and SBC) also include a penalty term which
is missing in the Log L test. Further, the Log L test can only be applied to nested models,
whereas AIC and SBC can also be applied to non-nested models. SBC has the advantage of
taking into consideration the degrees of freedom and likelihood ratio. SBC balances an
increase in likelihood with the number of parameters used to achieve that increase.
Also, SBC permits the judgment of several models before the selection of the most superior
model. However, in the case of the Log L test, a comparison can be made between a pair of
models and not multiple models at a point in time. Briefly, the drawbacks of a Log L test
which are its nesting requirement, its need for a paired comparison and the absence of
parameter importance, make it less desirable compared to SBC for its model selection
criteria (Posada and Buckley, 2004; Singer et al., 2014). Therefore, in the case of any
inconsistency in the results obtained by selection criteria, SBC will, finally, be considered for
model selection in the present study.

6. Efficacy of unconditional performance measurement models


6.1 Descriptive statistics
Table I presents descriptive statistics for mutual fund schemes and various benchmark
portfolios for two subperiods and the total period using daily data. Mutual fund schemes
exhibit an average excess return of 0.00096 (annualized[9] excess return of 24.05 percent) in
Subperiod I, − 0.00014 (annualized excess return of − 3.58 percent) in Subperiod II and
0.00017 (annualized excess return of 4.37 percent) in the total period. It can be seen that
mutual fund returns declined dramatically (in fact became negative) in the period since the
global financial crisis. In Subperiod I, 198 out of 199 mutual fund schemes exhibited positive
excess returns and 27 of them were able to outperform the average annualized
excess returns of 33.82 percent generated by market. On the other hand, in Subperiod II,
only 42 out of 237 schemes provided positive excess returns. In total, 195 schemes, however,
observed negative returns while, of those, 69 schemes reported higher negative returns
than that of the market generally, which generated an average annualized excess return of
− 4.02 percent. For the total period, 187 schemes generated positive excess returns but only
20 schemes outperformed the market index, which provided average annualized excess
returns of 12.62 percent.
Annualized average standard deviation of returns of mutual fund schemes remained
almost constant (around 18 percent) for three periods, irrespective of the returns earned by
schemes during different periods. For the total period, 15 funds exhibited higher risk than
the market generally in terms of standard deviation. However, their corresponding average
returns of − 2.25 percent were much lower than those of the market.
Average skewness for sample schemes was 0.17, ranging from − 1.28 to 43.11 over the
total period. The number of schemes exhibiting positive and negative skewness varied
drastically over the subperiods. Two out of 199 schemes demonstrated positive skewness in
Subperiod I which increased to 111 out of 237 in Subperiod II. However, for the total period,
65 schemes exhibited positive skewness and 172 schemes exhibited negative skewness.
Returns distributions of all schemes were leptokurtic for the total period as well as for the
JAMR Mean SD Skewness Kurtosis Jarque-Bera p-value
14,2
Subperiod I: April 2003-August 2007
MF schemes 0.00096 0.01137 −0.69187 12.89398 193,601.89070 0.00003
MKT 0.00135 0.01457 −1.07526 10.67860 2,834.85700 0.00000
SMB 0.00098 0.00561 −0.13963 3.93823 42.72290 0.00000
LMH 0.00035 0.00546 0.05981 4.08077 52.71426 0.00000
234 WML 0.00144 0.00721 0.10719 3.64077 20.35439 0.00004
RBT 0.00007 0.00002 0.52227 3.80202 77.32068 0.00000
CMA −0.00012 0.00669 0.46282 5.36544 287.65750 0.00000
WMR 0.00063 0.00786 0.51118 5.14347 251.43580 0.00000
QCFV 0.00155 0.00973 −0.14530 4.80631 149.22890 0.00000
Subperiod II: August 2007-March 2013
MF Schemes −0.00014 0.01112 0.19265 19.16798 118,171.57821 0.00000
MKT −0.00016 0.01712 0.16619 11.50306 4,109.41300 0.00000
SMB 0.00046 0.00542 −0.08520 6.40894 661.13230 0.00000
LMH 0.00002 0.00507 0.17452 4.29001 101.35340 0.00000
WML 0.00016 0.00867 −0.68078 8.77186 1,995.79200 0.00000
RBT 0.00007 0.00006 0.90498 2.46803 201.97140 0.00000
CMA 0.00008 0.00513 −0.00612 4.72688 169.24280 0.00000
WMR −0.00047 0.00699 −0.09017 4.09559 69.96353 0.00000
QCFV 0.00095 0.00869 −0.57618 6.43501 744.97160 0.00000
Total Period: April 2003-March 2013
MF Schemes 0.00017 0.01168 0.17049 26.17978 1,894,108.09814 0.00000
MKT 0.00051 0.01606 −0.26172 11.40651 7,188.93000 0.00000
SMB 0.00069 0.00551 −0.10549 5.22960 508.24800 0.00000
LMH 0.00017 0.00525 0.12455 4.19767 151.64200 0.00000
WML 0.00072 0.00808 −0.47878 7.69221 2,323.94800 0.00000
RBT 0.00007 0.00005 1.04193 3.60530 477.16400 0.00000
CMA −0.00001 0.00587 0.28571 5.51187 672.44600 0.00000
WMR 0.00002 0.00740 0.25263 4.82695 364.09400 0.00000
QCFV 0.00122 0.00916 −0.34067 5.62167 743.51900 0.00000
Notes: Average descriptive statistics for mutual fund schemes and various benchmark portfolios for two
subperiods and total period using daily data are shown below. For Subperiod I (April 1, 2003-August 8, 2007),
number of sample schemes is 199 and for Subperiod II (August 9, 2007-March 31, 2013), it is 237. MKT is
Table I. excess returns on market index (BSE 500 index), SMB is size factor, LMH represents value factor, WML is
Descriptive statistics stock momentum factor, RBT is excess return on government bond index, CMA represents investment factor,
for sample schemes WMR is profitability factor, QCFV represents firm quality factor. Descriptive statistics of sample schemes are
and factor portfolios available from authors on request

two subperiods. The Jarque-Bera ( JB) statistic was statistically significant at the 5 percent
level for all schemes confirming that their returns distributions were non-normal in nature.
Returns for the market index declined from strongly positive to negative for Subperiods
I and II. By contrast, standard deviation increased in period II. The market index exhibited
little skewness but strong leptokurtosis over the study period. The JB statistic was 7,188.93
suggesting that the distribution of the market index was not normal owing to leptokurtosis.
Descriptive statistics of other factors are also provided in the table. Other benchmark
portfolio returns also exhibited non-normality, as confirmed by the high JB statistic values.

6.2 Time series regressions


Table II presents a cross-sectional average of αs (intercepts) and βs (slope coefficients)[10] of
sample schemes for the total period for eight unconditional models. Two intercept terms, α
and α*, were reported. The former was for Subperiod I for 199 schemes and the latter was
JENSEN
Indian mutual
α α* β fund industry
Average of coefficient 0.00024 −0.00008 0.62818
t-statistic 9.67155 −6.63117 55.68680
Total observations 199 237 237
Number (+) 152 63 237
Number (−) 47 174 0
Significant (+) 68 8 237 235
Significant (−) 4 84 0
3M
α α* β γ
Average 0.00034 0.00003 0.62761 −0.35650
t-statistic 12.44749 1.99393 55.57970 −7.99265
Total observations 199 237 237 237
Number (+) 161 135 237 47
Number (−) 38 102 0 190
Significant (+) 93 7 237 11
Significant (−) 2 84 0 101
4M
α α* β γ δ
Average 0.00033 0.00003 0.62762 −0.35664 −0.33320
t-statistic 12.20730 2.14996 55.58131 −7.99601 −0.91179
Total observations 199 237 237 237 237
Number (+) 161 136 237 47 100
Number (−) 38 101 0 190 137
Significant (+) 95 8 237 19 29
Significant (−) 2 81 0 126 45
FF3F
α α* β s l
Average 0.00020 −0.00006 0.62826 0.18847 0.10132
t-statistic 8.54451 −4.48925 55.70920 14.27907 8.99238
Total observations 199 237 237 237 237
Number (+) 153 68 237 205 175
Number (−) 46 169 0 32 62
Significant (+) 59 8 237 183 131
Significant (−) 4 77 0 18 27
CARHART
α α* β s l w
Average 0.00016 −0.00005 0.62742 0.19155 0.12325 0.00070
t-statistic 7.55951 −4.06887 55.83221 14.72158 11.48223 0.14355
Total observations 199 237 237 237 237 237
Number (+) 144 75 237 205 190 143
Number (−) 55 162 0 32 47 94
Significant (+) 44 9 237 182 146 79
Significant (−) 4 68 0 20 21 40
ELTON
α α* β s l w b
Average 0.00020 −0.00003 0.62742 0.19155 0.12329 0.00061 −0.33942
t-statistic 6.95965 −1.64662 55.87773 14.75761 11.48176 0.12659 −1.66596
Total observations 199 237 237 237 237 237 237 Table II.
Number (+) 143 116 237 205 190 143 100 Model testing using
Number (−) 56 121 0 32 47 94 137 time series
regressions:
unconditional
(continued ) measures
JAMR Significant (+) 59 4 237 184 147 79 18
14,2 Significant (−) 4 66 0 19 21 40 38
FF5F
α α* β s l i p
Average 0.00020 −0.00006 0.62821 0.19158 0.10665 0.01064 −0.03976
t-statistic 8.27432 −5.00720 55.84190 14.06068 9.21704 2.24425 −8.92217
236 Total observations 199 237 237 237 237 237 237
Number (+) 149 69 237 203 174 130 53
Number (−) 50 168 0 34 63 107 184
Significant (+) 64 8 237 178 133 64 14
Significant (−) 4 80 0 20 23 33 117
FQUAL
α α* β s l i q
Average 0.00019 −0.00006 0.62822 0.19365 0.12084 0.00686 0.04793
t-statistic 8.40399 −4.43363 55.81879 14.29400 10.11320 1.50886 7.76248
Total observations 199 237 237 237 237 237 237
Number (+) 148 69 237 204 180 127 174
Number (−) 51 168 0 33 57 110 63
Significant (+) 61 8 237 179 138 53 96
Significant (−) 5 79 0 19 21 37 17
Notes: Cross-sectional average of αs (intercepts) and βs (slope coefficients) of sample schemes for total period
for eight unconditional models are shown below. We report two intercept terms α and α*. Former is for
Subperiod I for 199 schemes and latter is for Subperiod II for 237 schemes. The statistical significance of
Table II. parameters is tested at 5 percent level (two-tailed basis)

for Subperiod II for 237 schemes (which was equivalent to α in Subperiod I +/‒ the
coefficient of the dummy).
Using the JENSEN measure as a performance benchmark, the average slope of the
market factor for sample funds was 0.63. The β coefficient was found to be statistically
significant at the 5 percent level for all 237 schemes suggesting that excess returns of
growth funds in the Indian market were driven by excess market returns.
Using the 3M model, the average coefficient of the square of the market risk premium
(co-skewness measure) was − 0.36. The γ coefficient was found to be statistically significant
for about 47 percent of 237 schemes, out of which 90 percent of cases depicted a negative
coefficient. The co-skewness coefficient was statistically significant and exhibited a
negative sign which was consistent with the theory that investors prefer positive skewness
(Arditti, 1967; Kraus and Litzenberger, 1976; Scott and Horvath, 1980). It may be noted that,
since investors prefer positive skewness in their portfolios, negative market skewness is
viewed as risk. Given that, the market has a negatively skewed return distribution ( − 0.26),
on average, a mutual fund portfolio should have a positive skewness in order to hedge its
risk, so its co-skewness with the market should be negative, which is also observed in the
signs of the average coefficient of the square of a market risk premium. Our argument is
consistent with those of Friend and Westerfield (1980), Gregoriou et al. (2003) and Ding and
Shawky (2007), etc.
Using the 4M model, the average coefficient of the cube of market risk premium
(co-kurtosis measure), was − 0.33. The δ coefficient was found to be statistically significant
for about 31 percent of schemes, out of which about 61 percent of schemes portrayed a
negative coefficient. A negative sign for a co-kurtosis coefficient is consistent with the
theory that investors have negative preference for kurtosis (Scott and Horvath, 1980).
Market kurtosis was observed to be 11.40. Since the market is leptokurtic and high kurtosis
is considered as a negative incentive, so a mutual fund portfolio should have a low kurtosis
in order to hedge its risk and, as a result, its co-kurtosis with the market should be negative, Indian mutual
which is also observed in the signs of co-kurtosis measure[11]. fund industry
Using the FF3F model, the average coefficient of the size factor was 0.18. The size
coefficient was statistically significant for about 85 percent of schemes, out of which
91 percent of schemes demonstrated a positive coefficient. The average coefficient of value
factor was 0.10. The value coefficient was statistically significant for about 66 percent of
schemes, out of which about 83 percent exhibited a positive coefficient. The figures reflected 237
a fund manager’s investment style when they were engaging in size- and value-based
investment strategies by holding portfolios which were relatively more driven by small cap
stocks than by large cap stocks and by value stocks (low P/B stocks) than by growth stocks
(high P/B stocks).
Using the CARHART model, the average coefficient of momentum factor was 0.0007.
The returns on funds did not show a significant exposure to the momentum factor overall,
yet there was considerable evidence of a priced momentum factor on an individual basis.
The momentum coefficient was statistically significant in 50 percent of schemes, two-thirds
of which showed positive coefficients. Fund managers portray momentum-based investing
styles by preferring past winner stocks (over last 12 months) vis-à-vis past loser stocks.
Using the ELTON model as a performance benchmark, the average coefficient of bond index
factor was − 0.34. The average bond index coefficient was found to be statistically
insignificant. Considering individual regressions, 23 percent of 237 schemes showed a
statistically significant coefficient, out of which 68 percent of schemes were depicted as
having a negative coefficient. This negative sign exhibited interlinkages between equity and
bond prices. An increase in bond yield attracted new investors toward the bond market due
to higher returns and less risk. As investors prefer high-yield debts over equity, this shift in
preference of investors caused the equity market to decline and the prices of higher-yield
bonds to rise. Higher-yield debts in the bond market also led to a fall in demand for
lower-yield debts. As such, the prices of lower-yield debts began to decline in the secondary
market. Portfolios consisting of equity and/or lower-yield debts witnessed a fall in NAV
vis-à-vis portfolios comprised of higher-yield debts.
Using the FF5F model as a performance benchmark, the average coefficient of the
investment factor was 0.01. The investment coefficient was found to be statistically
significant in 41 percent of schemes, 66 percent of which showed positive coefficients.
The average coefficient of profitability factor was − 0.03. The profitability coefficient was
found to be statistically significant in 55 percent of schemes, 89.31 percent of which showed
a negative coefficient. It seems that the investment strategies of fund managers are focused
more on conservative investment stocks than on aggressive investment stocks and on
high-profitability stocks rather than on low-profitability stocks.
Using the FQUAL model, the average coefficient of firm quality factor was 0.04. The firm
quality coefficient was found to be statistically significant in 47 percent of schemes, out of
which about 85 percent demonstrated a positive coefficient reflecting that fund managers
are more exposed to low rather than high cash flow variability stocks.
It can be inferred from the above discussion that mutual fund managers in India devise
their investment strategies by concentrating more on small firms, value stocks, past winners
and more profitable firms which follow conservative investment policies and they exhibit
less cash flow variability. Also, co-skewness and co-kurtosis variables are priced in the
market in a higher order moments framework.

6.3 Model evaluation


Next, we tested how the performance benchmarks based on alternative asset pricing models
performed vis-à-vis each other. In other words, we attempted to identify the optimal
benchmark for mutual funds’ performance evaluation related to selectivity skills.
JAMR Table III shows cross-sectional averages of adjusted R2, AIC, SBC and Log L statistics for
14,2 sample schemes for each unconditional model using daily data.
The highest average adjusted R2 was found to be 0.82 for the ELTON model. The lowest
average AIC was observed for the FF5F Model while the lowest average SBC was found for
the CARHART model confirming the superiority of these models based on respective
evaluation criteria.
238 A Log L comparison was made considering each model as a base model. For instance, the
Log L of the Jensen measure was compared with that of the 3M model, the 4M model, etc.
Furthermore, the Log L of the 3M model was compared with that of the 4M model, the FF3F
model and so on. If twice the difference in the Log L of two models exceeded the tabulated χ2
(degrees of freedom) test statistic at a 5 percent level of significance, it indicated the
superiority of the new model over the old model (Otten and Bams, 2004). On a Log L basis,
the CARHART model seemed to offer the best performance evaluation benchmark.
Given the apparent inconsistency in model selection, we relied on SBC for those
reasons provided in Section 5 and concluded that CARHART gave the best performance
evaluation framework.
From a statistical viewpoint, it was concluded that amongst unconditional measures of
performance evaluation, the Carhart (1997) four-factor model was the optimal performance
benchmark for evaluating mutual fund performance in India. The findings were consistent
with those of Fletcher and Kihanda (2005), Otten and Bams (2004) and Lai and Lau (2010).

7. Efficacy of conditional performance measurement models


7.1 Time series regressions
Conditional time series regressions were estimated using two sets of IVs. Set I used three
public IVs, namely, DY, TB and TS. Set II incorporated DRS as a fourth IV in addition to the
above three variables.
Regressions for Set I were estimated covering the total period of the study from April
2003 to March 2013, and it was named, “Conditional without DRS.” Regressions for Set II
were estimated for the period from July 2004 to March 2013 and these were named,
“Conditional with DRS.” The reduced period under Set II was used because of the
non-availability of data about the DRS variable for India prior to July 2004. However, this set
was included to test the robustness of the results.
Table IV presents cross-sectional averages of αs (intercepts), βs (slope coefficients) and
Wald F-statistics of sample schemes for eight conditional models. Panels A and B provide
results for Set I – conditional without DRS and Set II – conditional with DRS, respectively.
7.1.1 Set I – conditional without DRS. In respect of all performance benchmarks, the
average Wald F-statistic was statistically significant. Results indicated that IVs were jointly
significant, implying that these variables captured the time-varying property of βs and
contained information that may have been useful for fund managers trying to predict
returns. Furthermore, considering individual regressions, the Wald F-statistic was found to
be statistically significant for all sample schemes for all models. It meant that βs of funds

JENSEN 3M 4M FF3F CARHART ELTON FF5F FQUAL


2
Adjusted R 0.80650 0.81025 0.81158 0.82491 0.82690 0.82698 0.82675 0.82671
Table III. AIC −7.99593 −8.01179 −8.01979 −8.08727 −8.09798 −8.09799 −8.09820 −8.09738
Model evaluation SBC −7.98744 −8.00025 −8.00519 −8.07267 −8.08032 −8.07727 −8.07749 −8.07666
based on selection Log L 7,651.75357 7,667.06106 7,675.20446 7,737.53173 7,748.94286 7,749.91979 7,749.82613 7,748.12461
criteria: unconditional Notes: It shows average values of adjusted R2, AIC, SBC and Log L for sample schemes based on model estimations
measures using daily data
Panel A: conditional without DRS
Indian mutual
fund industry
JENSEN
α α* βmkt Wald F
Average of
coefficient 0.00024 −0.00005 0.62463 844.43501
t-statistic 9.67432 −4.31077 54.75862
Total 239
observations 199 237 237
Number (+) 152 77 237
Number (−) 47 160 0
Significant (+) 75 7 237
Significant (−) 4 79 0
3M
α α* βmkt γmkt Wald F
Average 0.00034 0.00007 0.62524 −0.42856 891.96818
t-statistic 12.30922 4.35773 55.34821 −10.04793
Total
observations 199 237 237 237
Number (+) 159 156 237 44
Number (−) 40 81 0 193
Significant (+) 101 8 237 23
Significant (−) 2 76 0 155
4M
α α* βmkt γmkt δmkt Wald F
Average 0.00038 0.00006 0.62536 −0.41645 −1.37421 4677.48367
t-statistic 13.26422 3.73865 54.18993 −9.90600 −2.62957
Total
observations 199 237 237 237 237
Number (+) 163 153 237 41 89
Number (−) 36 84 0 196 148
Significant (+) 105 8 237 16 36
Significant (−) 2 81 0 145 78
FF3F
α α* βmkt smkt lmkt Wald F
Average 0.00018 −0.00004 0.62789 0.15892 0.05592 495.44660
t-statistic 7.87401 −2.83506 55.50971 12.43128 5.35502
Total
observations 199 237 237 237 237
Number (+) 148 83 237 194 152
Number (−) 51 154 0 43 85
Significant (+) 55 11 237 169 103
Significant (−) 5 68 0 28 57
CARHART
α α* βmkt smkt lmkt wmkt Wald F
Average 0.00016 −0.00003 0.62872 0.16013 0.08294 0.00890 429.59155
t-statistic 7.54574 −2.34046 56.07066 12.67489 8.09458 2.14959
Total
observations 199 237 237 237 237 237
Number (+) 145 86 237 195 165 156
Number (−) 54 151 0 42 72 81
Significant (+) 51 10 237 173 126 94 Table IV.
Significant (−) 5 64 0 26 41 33 Model testing using
time series
regressions:
(continued ) conditional measures
JAMR ELTON
14,2 α α* βmkt smkt lmkt wmkt bmkt Wald F
Average 0.00022 −0.00002 0.62883 0.15941 0.08355 0.00942 −0.17846 358.17173
t-statistic 7.82150 −1.22236 56.12242 12.70061 8.20252 2.26977 −0.91756
Total
observations 199 237 237 237 237 237 237
Number (+) 152 117 237 194 165 155 99
240 Number (−) 47 120 0 43 72 82 138
Significant (+) 64 8 237 172 128 97 19
Significant (−) 5 54 0 26 40 33 37
FF5F
α α* βmkt smkt lmkt imkt pmkt Wald F
Average 0.00017 −0.00003 0.62802 0.16170 0.06160 −0.01047 −0.02540 351.85616
t-statistic 7.71359 −2.44231 55.76812 12.44573 5.82778 −2.77590 −6.24814
Total
observations 199 237 237 237 237 237 237
Number (+) 147 91 237 192 152 125 66
Number (−) 52 146 0 45 85 112 171
Significant (+) 58 10 237 168 106 33 25
Significant (−) 4 69 0 29 57 48 113
FQUAL
α α* βmkt smkt lmkt imkt qmkt Wald F
Average 0.00018 −0.00003 0.62745 0.16434 0.07432 −0.01298 0.04705 344.55177
t-statistic 8.11767 −2.56069 55.68993 12.50389 6.84616 −3.54769 7.99684
Total
observations 199 237 237 237 237 237 237
Number (+) 147 83 237 196 155 118 183
Number (−) 52 154 0 41 82 119 54
Significant (+) 55 12 237 172 116 27 103
Significant (−) 4 70 0 27 45 56 20
Panel B: conditional with DRS
JENSEN
α α* βmkt Wald F
Average of
coefficient 0.00015 −0.00006 0.60601 829.87605
t-statistic 6.22293 −4.50607 48.38682
Total
observations 199 237 237
Number (+) 137 72 237
Number (−) 62 165 0
Significant (+) 36 7 237
Significant (−) 4 63 0
3M
α α* βmkt γmkt Wald F
Average 0.00025 0.00006 0.60689 −0.38342 875.61151
t-statistic 9.16728 3.41208 49.08711 −9.02651
Total
observations 199 237 237 237
Number (+) 153 137 237 48
Number (−) 46 100 0 189
Significant (+) 76 10 237 22
Significant (−) 4 62 0 149

Table IV. (continued )


4M Indian mutual
α α* βmkt γmkt δmkt Wald F fund industry
Average 0.00028 0.00002 0.60775 −0.34028 −1.43266 5,508.72887
t-statistic 9.73223 1.15664 48.12645 −8.34459 −3.19569
Total
observations 199 237 237 237 237
Number (+) 158 115 237 53 86
Number (−) 41 122 0 184 151 241
Significant (+) 80 10 237 21 38
Significant (−) 2 82 0 118 80
FF3F
α α* βmkt smkt lmkt Wald F
Average 0.00009 −0.00004 0.61008 0.15073 0.04826 464.04032
t-statistic 4.15502 −3.08179 49.15010 11.82293 4.67624
Total
observations 199 237 237 237 237
Number (+) 127 84 237 190 139
Number (−) 72 153 0 47 98
Significant (+) 28 10 237 166 89
Significant (−) 5 40 0 31 57
CARHART
α α* βmkt smkt lmkt wmkt Wald F
Average 0.00006 −0.00004 0.61207 0.15444 0.07792 0.01216 398.92957
t-statistic 2.75789 −2.84852 50.10128 12.38279 7.80171 2.84469
Total
observations 199 237 237 237 237 237
Number (+) 117 87 237 193 165 163
Number (−) 82 150 0 44 72 74
Significant (+) 17 11 237 167 119 112
Significant (−) 7 27 0 29 36 37
ELTON
α α* βmkt smkt lmkt wmkt bmkt Wald F
Average 0.00010 −0.00007 0.61199 0.15329 0.07863 0.01278 0.39335 330.71627
t-statistic 3.88266 −3.93656 50.12661 12.38253 7.94578 2.96806 2.17647
Total
observations 199 237 237 237 237 237 237
Number (+) 125 96 237 193 165 164 126
Number (−) 74 141 0 44 72 73 111
Significant (+) 28 6 237 166 118 113 23
Significant (−) 9 39 0 28 35 37 21
FF5F
α α* βmkt smkt lmkt imkt pmkt Wald F
Average 0.00008 −0.00003 0.61003 0.15397 0.05668 −0.01173 −0.02185 335.59168
t-statistic 3.73264 −2.11064 49.25425 11.84140 5.41102 −3.12829 −5.38437
Total
observations 199 237 237 237 237 237 237
Number (+) 126 95 237 187 142 114 72
Number (−) 73 142 0 50 95 123 165
Significant (+) 27 12 237 167 98 28 24
Significant (−) 5 39 0 31 54 52 104

(continued ) Table IV.


JAMR FQUAL
14,2 α α* βmkt smkt lmkt imkt qmkt Wald F
Average 0.00009 −0.00003 0.60907 0.15604 0.06616 −0.01295 0.04344 326.72684
t-statistic 4.11973 −2.57044 49.27053 11.88253 6.16283 −3.54658 6.86246
Total
observations 199 237 237 237 237 237 237
Number (+) 129 88 237 189 145 110 172
242 Number (−) 70 149 0 48 92 127 65
Significant (+) 25 9 237 165 101 22 94
Significant (−) 5 37 0 31 47 53 19
Notes: Cross-sectional average of αs (intercepts), βs (slope coefficients) and Wald F-statistics of sample
schemes for eight conditional models are shown below. Panels A and B provide results for Set I – conditional
without DRS and Set II – conditional with DRS, respectively. Results related to information variables are
Table IV. available from authors on request

were not constant but time-varying and it signified the presence of dynamic strategies
employed by fund managers as responses to economic or market conditions.
7.1.2 Set II – conditional with DRS. The average Wald F-statistic was found to be
statistically significant. Also, the Wald F-statistic, corresponding to individual regressions,
was observed to be statistically significant indicating the joint significance of IVs.
Broadly, these results suggested that publicly available IVs have been actively
integrated by mutual fund managers in India for formulating their investment strategies
in predicting expected returns and deciding the risk exposures of fund portfolios. Hence,
the conditional performance evaluation approach should be employed while evaluating
the performance of mutual funds to account for the investment strategy devised on the
basis of public IVs.

7.2 Model evaluation


Panels A and B of Table V provide average values of adjusted R2, AIC, SBC and Log L for
sample schemes for each model of Set I – conditional without DRS and Set II – conditional
with DRS, respectively.
7.2.1 Set I – conditional without DRS. A choice of the best model based on the highest
average adjusted R2 was the ELTON model. The lowest average AIC and SBC were
reported for the CARHART model. Further, the Log likelihood comparison indicated that
the ELTON model was statistically superior to other models.

JENSEN 3M 4M FF3F CARHART ELTON FF5F FQUAL

PANEL A: conditional without DRS


Adjusted R2 0.81910 0.82388 0.82785 0.83781 0.84068 0.84094 0.84005 0.84038
AIC −8.10690 −8.13060 −8.15235 −8.20633 −8.22326 −8.22267 −8.21739 −8.21808
SBC −8.08896 −8.10023 −8.10956 −8.16354 −8.16805 −8.15504 −8.14976 −8.15045
Log L 7,604.19583 7,630.86626 7,655.84516 7,702.94851 7,723.01547 7,726.54791 7,720.59694 7,720.86453
PANEL B: conditional with DRS
Adjusted R2 0.82899 0.83370 0.83644 0.84736 0.84993 0.85013 0.85035 0.85006
Table V. AIC −8.26997 −8.29355 −8.30972 −8.37186 −8.38853 −8.38712 −8.38759 −8.38395
Model evaluation SBC −8.24785 −8.25514 −8.25501 −8.31715 −8.31752 −8.29981 −8.30029 −8.29665
based on selection Log L 7,179.92734 7,205.19766 7,223.57149 7,276.83924 7,296.49591 7,300.25251 7,299.85151 7,296.18023
criteria – conditional Notes: Panels A and B provide average values of adjusted R2, AIC, SBC and Log L for sample schemes for
measures Set I – conditional without DRS and Set II – conditional with DRS variable, respectively
Given the apparent inconsistency in model selection based on SBC, it was concluded that Indian mutual
CARHART was the best performance evaluation framework. fund industry
Hence, from a statistical standpoint, it was concluded that in a conditional setting, where
the IVs used were lagged DY, TB and TS, the Carhart (1997) four-factor asset pricing model
was the optimal performance benchmark against which to measure mutual fund performance.
7.2.2 Set II – conditional with DRS. The highest average adjusted R2 was obtained for
the FF5F model while the lowest average AIC and SBC were observed for the CARHART 243
model. Log L testing showed that the ELTON and FF5F models were better than other
remaining models. Owing to a result disparity, we followed the SBC criterion which showed
that the CARHART model was the best performance benchmark. That is, in a conditional
setting where the IVs used were lagged DY, TB, TS and DRS, the Carhart (1997) four-factor
model was selected as the best performance benchmark.
Broadly, based on both sets of time series regressions, it was concluded that the Carhart
(1997) four-factor model was the best performance benchmark irrespective of the public IVs used.

8. Unconditional vs conditional models: a comparison


By means of model selection criteria, this section compares unconditional models with their
conditional counterparts using Conditional without DRS (Set I), since the numbers of
observations for both these sets of models matched with each other.
The average adjusted R2 for all conditional models was observed to be higher than that
of unconditional models indicating that model specification in conditional settings was
better than it was in unconditional settings. These results were consistent with those of
Ferson and Schadt (1996), Roy and Deb (2004), Aragon and Ferson (2006), Moreno and
Rodriguez (2009) and Roy (2015a, b).
Average AICs and SBCs for all conditional models were found to be lower than for
unconditional models confirming the superiority of conditional models over unconditional
models. However, on the basis of the average Log L, conditional models were not found to be
statistically different from their unconditional counterparts.
It was concluded that performance benchmarks based on conditional approaches that
recognize the time-varying properties of βs, had an edge over the performance benchmarks
based on unconditional approaches, assuming constant βs.

9. Concluding observations
This study has examined the efficacy of performance benchmarks based on alternative
asset pricing models for the mutual fund industry in India. Using daily observations for a
sample of 237 open-ended Indian equity (growth) schemes from April 2003 to March 2013,
both unconditional and conditional versions of eight performance models were employed,
namely, the Jensen (1968) measure, the three-moment model, the four-moment model, the
Fama and French (1993) three-factor model, Carhart’s (1997) four-factor model, the Elton
et al. (1999) five-index model, the Fama and French (2015) five-factor model and the firm
quality five-factor model.
The major findings of our study are: amongst the unconditional versions, the
Carhart (1997) four-factor model seemed to offer the most superior performance benchmark
based on the selected model selection criteria. The Carhart (1997) model again outperformed
all other models as a conditional framework. Three, conditional models outperformed their
unconditional counterparts. Thus, it was concluded that the conditional Carhart (1997)
four-factor model was the most appropriate performance benchmark for the Indian mutual
fund industry.
These findings have implications for mutual fund investors, fund managers, regulators
and the academic community.
JAMR Investors are confronted with the complicated task of selecting better performing funds
14,2 out of a range of investment options available in the market. Hence, they need to be
acquainted with the most superior performance benchmark, which will facilitate their
decision-making. From an investor’s perspective, the study recommends the use of a
multifactor performance benchmark, namely, the Carhart (1997) four-factor model, that
takes into account the investment-style characteristics of fund managers. A significant α
244 generated over and above the return estimated using the Carhart (1997) model reflects the
true stock-picking skills of fund managers and it is, therefore, worth paying an active
management fee for this. However, if most of an α is ascribed to common styles of investing
in small firms, value stocks and past winner stocks, then an investor will have no incentive
to pay excessive fees to mutual fund managers for their stock selection. In addition, the
ability of conditional models to identify the behavior of βs, allows investors to examine
the dynamic investment styles of fund managers. Furthermore, the findings clarifying
the investment strategies of mutual fund managers in India offer valuable insights to
investors making judicious investment decisions. Managers of open-ended, equity-oriented
Indian mutual fund schemes, with growth as their objective, devise their investment
strategies by concentrating more on small firms, value stocks, past winners and more
profitable firms which follow conservative investment policies and they exhibit less cash
flow variability. In addition, it was reflected how managers of growth funds hold portfolios
that adequately account for the risks associated with the higher order moments, namely,
skewness and kurtosis, as the signs exhibited by co-skewness and co-kurtosis Variables
were found consistent with the theory that risk-averse investors prefer positive skewness
and dislike high kurtosis in their portfolios.
The Carhart (1997) model should also serve as a benchmark for fund managers to
monitor their investment actions as they strive to achieve a certain level of performance.
Funds may use the Carhart (1997) model as a risk-adjusted performance measure for
managerial evaluation. Also, any reporting of risk-adjusted returns should be based on this
model so that better performing managers can be distinguished accurately from less
well-performing managers. Furthermore, the selection of managers by mutual fund houses
should be made prudently by tracking their past performance using those aforementioned
performance benchmarks, as an evaluation based on any other performance benchmark
may lead to improper managerial selection. Furthermore, it may not be cost-effective to
terminate such an advisory relationship and switch to a new manager at a later stage.
The Carhart (1997) model should serve as a mode for the pricing assets and, thus, be
useful for market regulators who aim for fair trades and efficient markets. Construction of
suitable benchmarks is recommended for enabling investors and managers to undertake
meaningful evaluations of the performance of funds.
Currently, Indian indices limit themselves to the market-capitalization-based and
sector-specific indices. The benchmarks, specifically incorporating size, value and
momentum factors in addition to market factors to evaluate the mutual funds, may have
to be evolved for the purpose of the accurate evaluation of funds. Thus, stock exchanges,
namely, the NSE and the BSE, and credit rating agencies, namely, CRISIL and ICRA should
construct such indices and make them available in the public domain to be used for the
purpose of benchmarking.
Besides, these findings being of certain importance for researchers, since the existing
literature on the assessment of unconditional and conditional performance measures is
meager for emerging markets including India, this study contributes to the literature of
mutual fund performance evaluation by suggesting the optimal performance benchmark for
the Indian mutual fund industry.
The study also highlights further research issues that need to be explored: For instance
what should be the optimal performance benchmark in the cases of other types of fund
schemes such as balanced funds, income funds, sector funds, global funds, etc., given that Indian mutual
the present study covers only growth funds? Could a macroeconomic, factor-based fund industry
performance model that was consistent with arbitrage pricing theory (APT framework)
provide more optimal performance benchmarks? Could certain fund- and managerial-based
characteristics be used to predict fund performance? Are our results country-specific or do
they also hold for other emerging markets?
Further research on the subject is recommended, given its economic as well as its 245
academic importance.

Notes
1. Momentum was a phenomenon first recorded by Jegadeesh and Titman (1993), which implies that
past winners continue to be future winners whilst past losers continue to be future losers over the
subsequent from three to 12 months of a portfolio holding period.
2. Sehgal and Subramaniam (2012) documented how more profitable firms tended to have higher
dividend payouts and were therefore perceived to be less risky by investors resulting in lower
returns for such firms.
3. Prices deviate from NAV owing to market pessimism or optimism even around the maturity
dates of schemes.
4. In India, companies follow April-March as their financial years for reporting purposes.
The sample period in the present study also followed the financial year concept.
5. Sehgal and Jain (2015) stated that the classification of stocks in 2X3 portfolios resulted in a higher
correlation between size and value factors than that based on 2X2 portfolios. 2X2 portfolios are
formed to avoid multicollinearity between factors.
6. While constructing investment and profitability factors, the present study divided portfolios into
quintiles unlike Fama and French (2015) who performed 2X2X2X2 sorts for size, value,
investment and profitability factors. For details, see Fama and French (2015).
7. Growth in total assetst ¼ (Total assets at the end of year t − Total assets at the end of year t − 1/
Total assets at the end of year t − 1).
8. Wherever the correlation between two regressors is greater than 0.50, an auxiliary regression is
run using the orthogonalizing procedure of Giliberto (1985) which considers one of the two
regressors as a dependent variable and the other as an independent variable. The residual series
generated from such regression can be taken as a new variable for further estimations and
analysis.
9. Following Sehgal and Jhanwar (2008), daily data were annualized assuming 250 working days in
a year.
10. The β (slope coefficient) was a cross-sectional average from a series of coefficient estimates and
the t-statistic was an average slope divided by its time series standard error. See Fama and
French (1992) for reporting estimates.
11. The argument concerning co-kurtosis is extended from and consistent with that of co-skewness
as proposed by Friend and Westerfield (1980).

References
Agudo, L.F., Magallon, M.V. and Sarto, J.L. (2006), “Evaluation of performance and conditional
information: the case of Spanish mutual funds”, Applied Financial Economics, Vol. 16 No. 11,
pp. 803-817.
Ajibola, A., Kunle, O.A. and Prince, C.N. (2015), “Empirical proof of the CAPM with higher order
co-moments in Nigerian stock market: the conditional and unconditional based tests”,
Journal of Applied Finance & Banking, Vol. 5 No. 1, pp. 151-162.
JAMR Anand, S. and Murugaiah, V. (2006), “Analysis of components of investment performance – an
14,2 empirical study of mutual funds in India”, 10th Indian Institute of Capital Markets Conference,
December 18, available at: http://ssrn.com/abstract=961999; http://dx.doi.org/10.2139/ssrn.961999
(accessed July 2012).
Aragon, G.O. and Ferson, W.E. (2006), “Portfolio performance evaluation”, Foundations and Trends in
Finance, Vol. 2 No. 2, pp. 83-190.
246 Arditti, F.D. (1967), “Risk and the required return on equity”, The Journal of Finance, Vol. 22 No. 1,
pp. 19-36.
Arditti, F.D. (1971), “Another look at mutual fund performance”, Journal of Financial and Quantitative
Analysis, Vol. 6 No. 3, pp. 909-912.
Barone Adesi, G., Gagliardini, P. and Urga, G. (2004), “Testing asset pricing models with coskewness”,
Journal of Business & Economic Statistics, Vol. 22 No. 4, pp. 474-485.
Barua, S.K. and Varma, J.R. (1991), “Master shares: a bonanza for large Investors”, Vikalpa, Vol. 16
No. 1, pp. 29-34.
Bessler, W., Drobetz, W. and Zimmermann, H. (2009), “Conditional performance evaluation for German
equity mutual funds”, The European Journal of Finance, Vol. 15 No. 3, pp. 287-316.
Bollen, N.P. and Busse, J.A. (2001), “On the timing ability of mutual fund managers”, The Journal of
Finance, Vol. 56 No. 3, pp. 1075-1094.
Brockett, P.L. and Kahane, Y. (1992), “Risk, return, skewness and preference”, Management Science,
Vol. 38 No. 6, pp. 851-866.
Carhart, M. (1997), “On persistence in mutual fund performance”, The Journal of Finance, Vol. 52 No. 1,
pp. 57-82.
Cavanaugh, J.E. (2009), 171: 290 Model Selection, Lecture VI: The Bayesian Information Criterion,
Department of Biostatistics, Department of Statistics and Actuarial Science, University of
Iowa, Iowa.
Chandel, K. and Verma, O.P. (2005), “Managing mutual fund investments in the era of change”,
The ICFAI Journal of Applied Finance, Vol. 11 No. 9, pp. 56-67.
Chander, R. (2005), “Empirical investigation on the investment manager’s stock selection abilities:
the Indian experience”, The ICFAI Journal of Applied Finance, Vol. 11 No. 7, pp. 5-20.
Chaudhry, S. and Shakeel, M. (2014), “Performance evaluation of top performing mutual fund
managers: an analytical study from India”, International Journal of Research in Commerce,
Economics and Management, Vol. 4 No. 9, pp. 71-76.
Christie, D.R. and Chaudhry, M. (2001), “Coskewness and cokurtosis in futures markets”, Journal of
Empirical Finance, Vol. 8 No. 1, pp. 55-81.
Christopherson, J.A., Ferson, W.E. and Glassman, D.A. (1998), “Conditioning manager alphas on
economic information: another look at the persistence of performance”, Review of Financial
Studies, Vol. 11 No. 1, pp. 111-142.
Cortez, M.C., Silva, F. and Areal, N. (2009), “The performance of European socially responsible funds”,
Journal of Business Ethics, Vol. 87 No. 4, pp. 573-588.
Dahlquist, M., Engstrom, S. and Soderlind, P. (2000), “Performance and characteristics of Swedish
mutual funds”, Journal of Financial and Quantitative Analysis, Vol. 35 No. 3, pp. 409-423.
Dhar, J. (2013), “Stock selection skills of Indian mutual fund managers during 2000-2012”, IOSR Journal
of Business and Management, Vol. 10 No. 1, pp. 79-87.
Ding, B. and Shawky, H.A. (2007), “The performance of hedge fund strategies and the asymmetry of
return distributions”, European Financial Management, Vol. 13 No. 2, pp. 309-331.
Doan, P., Lin, C.T. and Zurbruegg, R. (2010), “Pricing assets with higher moments: evidence from the
Australian and US stock markets”, Journal of International Financial Markets, Institutions and
Money, Vol. 20 No. 1, pp. 51-67.
Elton, E., Gruber, M. and Blake, C. (1999), “Common factors in active and passive portfolios”, European Indian mutual
Finance Review, Vol. 3 No. 1999, pp. 53-78. fund industry
Elton, E., Gruber, M., Das, S. and Hlavka, M. (1993), “Efficiency with costly information:
a reinterpretation of evidence from managed portfolios”, Review of Financial Studies, Vol. 6
No. 1, pp. 1-22.
Fama, E.F. (1970), “Efficient capital markets: a review of theory and empirical work”, The Journal of
Finance, Vol. 25 No. 2, pp. 383-417. 247
Fama, E.F. and French, K.R. (1992), “The cross‐section of expected stock returns”, The Journal of
Finance, Vol. 47 No. 2, pp. 427-465.
Fama, E.F. and French, K.R. (1993), “Common risk factors in the returns on stocks and bonds”,
Journal of Financial Economics, Vol. 33 No. 1, pp. 3-56.
Fama, E.F. and French, K.R. (1996), “Multi-factor explanations of asset-pricing anomalies”, Journal of
Finance, Vol. 51 No. 1, pp. 55-84.
Fama, E.F. and French, K.R. (2015), “A five-factor asset pricing model”, Journal of Financial Economics,
Vol. 116 No. 1, pp. 1-22.
Fang, H. and Lai, T. (1997), “Co-kurtosis and capital asset pricing”, Financial Review, Vol. 32 No. 2,
pp. 293-307.
Ferson, W.E. and Schadt, R.W. (1996), “Measuring fund strategy and performance in changing
economic conditions”, The Journal of Finance, Vol. 51 No. 2, pp. 425-461.
Filardo, A., George, J., Loretan, M., Ma, G., Munro, A., Shim, I. and Zhu, H. (2010), “The international
financial crisis: timeline, impact and policy responses in Asia and the Pacific”, BIS Papers,
Vol. 52, pp. 21-82, available at: www.bis.org/publ/bppdf/bispap52c.pdf
Fletcher, J. (2002), “Examination of conditional asset pricing in UK stock returns”, Financial Review,
Vol. 37 No. 3, pp. 447-468.
Fletcher, J. and Forbes, D. (2002), “UK unit trust performance: does it matter which benchmark or
measure is used?”, Journal of Financial Services Research, Vol. 21 No. 3, pp. 195-218.
Fletcher, J. and Kihanda, J. (2005), “An examination of alternative CAPM-based models in UK stock
returns”, Journal of Banking & Finance, Vol. 29 No. 12, pp. 2995-3014.
Friend, I. and Westerfield, R. (1980), “Co‐skewness and capital asset pricing”, The Journal of Finance,
Vol. 35 No. 4, pp. 897-913.
Gallefoss, K., Hansen, H.H., Haukaas, E.S. and Molnár, P. (2015), “What daily data can tell us about
mutual funds: evidence from Norway”, Journal of Banking & Finance, Vol. 55 No. 2015,
pp. 117-129.
Gardijan, M. and Skrinjaric, T. (2015), “Estimating investor preferences towards portfolio
return distribution in investment funds”, Croatian Operational Research Review, Vol. 6 No. 1,
pp. 1-16.
Giliberto, M. (1985), “Interest rate sensitivity in the common stocks of financial intermediaries:
a methodological note”, Journal of Financial and Quantitative Analysis, Vol. 20 No. 1,
pp. 123-126.
Goetzmann, W.N., Ingersoll, J. and Ivković, Z. (2000), “Monthly measurement of daily timers”,
Journal of Financial and Quantitative Analysis, Vol. 35 No. 3, pp. 257-290.
Gregoriou, G.N., Karavas, V.N. and Rouah, F.D. (Eds) (2003), Hedge Funds: Strategies, Risk Assessment,
and Returns, Beard Books, Washington, DC.
Grinblatt, M. and Titman, S. (1989), “Mutual fund performance: an analysis of quarterly portfolio
holdings”, The Journal of Business, Vol. 62 No. 3, pp. 393-416.
Gujarati, D.N. (2004), Basic Econometrics, Tata McGraw-Hill Publishing Company, New Delhi.
Harvey, C.R. and Siddique, A. (2000), “Conditional Skewness in asset pricing tests”, The Journal of
Finance, Vol. LV No. 3, pp. 1263-1295.
JAMR Hasan, M.Z. and Kamil, A.A. (2014), “Contribution of co-skewness and co-kurtosis of the higher
14,2 moment CAPM for finding the technical efficiency”, Economics Research International, pp. 1-9.
Hasan, M.Z., Kamil, A.A., Mustafa, A. and Baten, M.A. (2013), “An empirical analysis of higher moment
capital asset pricing model for Bangladesh stock market”, Modern Applied Science, Vol. 7 No. 5,
pp. 11-21.
Hung, D.C.H., Shackleton, M. and Xu, X. (2004), “CAPM, higher co-moment and factor models of UK
248 stock returns”, Journal of Business Finance and Accounting, Vol. 31 Nos 1-2, pp. 87-112.
Hwang, S. and Satchell, S.E. (1999), “Modelling emerging market risk Premia using higher moments”,
Journal of Finance and Economics, Vol. 4 No. 4, pp. 271-296.
Ippolito, R.A. (1989), “Efficiency with costly information: a study of mutual fund performance,
1965-1984”, The Quarterly Journal of Economics, Vol. 104 No. 1, pp. 1-23.
Jagannathan, R. and Wang, Z. (1996), “The conditional CAPM and the cross-section of expected
returns”, Journal of Finance, Vol. 51 No. 1, pp. 3-53.
Javid, A. (2009), “Test of higher moment capital asset pricing model in case of Pakistani
equity market”, European Journal of Economics, Finance and Administrative Sciences, Vol. 15
No. 2009, pp. 144-162.
Jayadev, M. (1996), “Mutual fund performance: an analysis of monthly returns”, Finance India, Vol. 10
No. 1, pp. 73-84.
Jean, W.H. (1971), “The extension of portfolio analysis to three or more parameters”, Journal of
Financial and Quantitative Analysis, Vol. 6 No. 1, pp. 505-515.
Jegadeesh, N. and Titman, S. (1993), “Returns to buying winners and selling losers: implications for
stock market efficiency”, The Journal of Finance, Vol. 48 No. 1, pp. 65-91.
Jensen, M.C. (1968), “The performance of mutual funds in the period 1945-1964”, The Journal of
Finance, Vol. 23 No. 2, pp. 389-416.
Kadane, J.B. and Lazar, N.A. (2004), “Methods and criteria for model selection”, Journal of the American
Statistical Association, Vol. 99 No. 465, pp. 279-290.
Kaur, I. (2013), “Performance, timing and selectivity skills of Indian Equity Mutual funds: an empirical
approach”, Researchers World, Vol. 4 No. 4, pp. 87-94.
Kraus, A. and Litzenberger, R.H. (1976), “Skewness preference and the valuation of risk assets”,
The Journal of Finance, Vol. 31 No. 4, pp. 1085-1100.
Lai, M.M. and Lau, S.H. (2010), “Evaluating mutual fund performance in an emerging Asian economy:
The Malaysian experience”, Journal of Asian Economics, Vol. 21 No. 4, pp. 378-390.
Lance, C.E. (1988), “Residual centering, exploratory and confirmatory moderator analysis, and
decomposition of effects in path models containing interactions”, Applied Psychological
Measurement, Vol. 12 No. 2, pp. 163-175.
Lee, S.L. (1999), “The conditional performance of UK property funds”, paper presented at the
15th Annual American Real Estate Society Meeting.
Levy, H. (1969), “A utility function depending on the first three moments”, The Journal of Finance,
Vol. 24 No. 4, pp. 715-719.
Lin, B.H. and Wang, J.M. (2003), “Systematic skewness in asset pricing: an empirical examination of the
Taiwan stock market”, Applied Economics, Vol. 35 No. 17, pp. 1877-1887.
Lin, B.H. and Wang, J.M. (2004), “Asset pricing with higher moments: empirical evidence from the
Taiwan stock market”, in Lee, C.F. (Ed.), Advances in Quantitative Analysis of Finance and
Accounting, Vol. 1, World Scientific Publishing Co., pp. 153-170.
Lintner, J. (1965), “The valuation of risk assets and the selection of risky investments in stock
portfolios and capital budgets”, The Review of Economics and Statistics, Vol. 47 No. 1,
pp. 13-37.
Liow, K.H. and Chan, L.C. (2005), “Co‐skewness and co‐kurtosis in global real estate securities”,
Journal of Property Research, Vol. 22 Nos 2-3, pp. 163-203.
Malkiel, B.G. (1995), “Returns from investing in equity mutual funds 1971 to 1991”, The Journal of Indian mutual
Finance, Vol. 50 No. 2, pp. 549-572. fund industry
Merton, R.C. (1973), “An intertemporal capital asset pricing model”, Econometrica: Journal of the
Econometric Society, Vol. 41 No. 5, pp. 867-887.
Messis, P., Iatridis, G. and Blanas, G. (2007), “CAPM and the efficacy of higher moment CAPM in the
Athens stock market: an empirical approach”, International Journal of Applied Economics, Vol. 4
No. 1, pp. 60-75. 249
Miller, M. and Modigliani, F. (1961), “Dividend policy, growth, and the valuation of shares”, Journal of
Business, Vol. 34 No. 4, pp. 411-433.
Moreno, D. and Rodriguez, R. (2005), “Performance evaluation considering the coskewness: a stochastic
discount factor framework”, Managerial Finance, Vol. 32 No. 4, pp. 375-392.
Moreno, D. and Rodriguez, R. (2009), “The value of coskewness in mutual fund performance
evaluation”, Journal of Banking and Finance, Vol. 33 No. 9, pp. 1664-1676.
Naz, S., Mustafa, A.U., Mukhtar, A. and Nawaz, S. (2015), “Risk adjusted performance evaluation of
balanced mutual fund schemes in Pakistan”, European Journal of Business and Management,
Vol. 7 No. 1, pp. 179-187.
Neto, N. (2014), “Do Portuguese mutual funds display forecasting skills? A study of selectivity and
market timing ability”, Master Dissertation in Finance, School of Economics and Management,
University of Porto.
Newey, W.K. and West, K.D. (1987), “Hypothesis testing with efficient method of moments estimation”,
International Economic Review, Vol. 28 No. 3, pp. 777-787.
Otten, R. and Bams, D. (2004), “How to measure mutual fund performance: economic versus statistical
relevance”, Accounting & Finance, Vol. 44 No. 2, pp. 203-222.
Pandey, A. and Sehgal, S. (2017), “Volatility effect and the role of firm quality factor in returns:
evidence from Indian stock market”, IIMB, Management Review, doi: 10.1016/j.iimb.2017.01.001.
Posada, D. and Buckley, T.R. (2004), “Model selection and model averaging in Phylogenetics:
advantages of Akaike information criterion and Bayesian approaches over likelihood ratio
tests”, Systematic Biology, Vol. 53 No. 5, pp. 793-808.
Ranaldo, A. and Favre, L. (2005), “Hedge fund performance and higher-moment market models”,
The Journal of Alternative Investments, Vol. 8 No. 3, pp. 37-51.
Ross, S.A. (1976), “The arbitrage theory of capital asset pricing”, Journal of Economic Theory, Vol. 13
No. 3, pp. 341-360.
Roy, B. and Deb, S.S. (2004), “The conditional performance of Indian Mutual Funds: an empirical
study”, available at: http://ssrn.com/abstract=593723; http://dx.doi.org/10.2139/ssrn.593723
(accessed July 24, 2014).
Roy, S. (2015a), “An empirical study between traditional and conditional mutual fund performance:
Indian evidence”, Indian Journal of Accounting, Vol. 47 No. 1, pp. 38-59.
Roy, S. (2015b), “Conditional selectivity performance of Indian mutual fund schemes: an empirical
study”, Management Science Letters, Vol. 5 No. 6, pp. 577-590.
Roy, S. (2016), “Another look in conditioning Alphas on economic information: Indian evidence”, Global
Business Review, Vol. 17 No. 1, pp. 191-213.
Rubinstein, M.E. (1973), “The fundamental theorem of parameter-preference security valuation”,
Journal of Financial and Quantitative Analysis, Vol. 8 No. 1, pp. 61-69.
Sawicki, J. and Ong, F. (2000), “Evaluating managed fund performance using conditional measures:
Australian evidence”, Pacific-Basin Finance Journal, Vol. 8 No. 3, pp. 505-528.
Scott, R.C. and Horvath, P.A. (1980), “On the direction of preference for moments of higher order than
the variance”, The Journal of Finance, Vol. 35 No. 4, pp. 915-919.
Sears, R.S. and Wei, K.C. (1985), “Asset pricing, higher moments, and the market risk premium: a note”,
The Journal of Finance, Vol. 40 No. 4, pp. 1251-1253.
JAMR Sehgal, S. and Jain, K. (2015), “Dissecting sources of price momentum: evidence from India”,
14,2 International Journal of Emerging Markets, Vol. 10 No. 4, pp. 801-819.
Sehgal, S. and Jhanwar, M. (2008), “On stock selection skills and market timing abilities of mutual fund
managers in India”, International Research Journal of Finance and Economics, Vol. 15 No. 1,
pp. 307-317.
Sehgal, S. and Subramaniam, S. (2012), “An empirical investigation of the profitability anomaly in the
Indian stock market”, Asian Journal of Finance & Accounting, Vol. 4 No. 2, pp. 347-362.
250
Shanmugham, R. and Zabiulla (2011), “Return-based performance analysis of selected equity mutual
funds schemes in India – an empirical study”, International Journal of Research in Computer
Application & Management, Vol. 1 No. 1, pp. 113-119.
Sharpe, W.F. (1964), “Capital asset prices: a theory of market equilibrium under conditions of risk”,
Journal of Finance, Vol. 19 No. 3, pp. 425-442.
Sharpe, W.F. (1966), “Mutual fund performance”, The Journal of Business, Vol. 39 No. 1, pp. 119-138.
Singer, P., Helic, D., Taraghi, B. and Strohmaier, M. (2014), “Detecting memory and structure in human
navigation patterns using Markov chain models of varying order”, PLoS ONE, Vol. 9 No. 7,
p. e102070, available at: http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0102070
(accessed September 15, 2015).
Swinkels, L. and Rzezniczak, P. (2009), “Performance evaluation of Polish mutual fund managers”,
International Journal of Emerging Markets, Vol. 4 No. 1, pp. 26-42.
Titman, S., Wei, K.J. and Xie, F. (2004), “Capital investments and stock returns”, Journal of Financial
and Quantitative Analysis, Vol. 39 No. 4, pp. 677-700.
Treynor, J.L. (1965), “How to rate management of investment funds”, Harvard Business Review, Vol. 43
No. 1, pp. 63-75.
Tsiang, S.C. (1972), “The rationale of the mean-standard deviation analysis, skewness preference, and
the demand for money”, The American Economic Review, Vol. 62 No. 3, pp. 354-371.
Walkshausl, C. (2013), “The high returns to low volatility stocks are actually a premium on high quality
firms”, Review of Financial Economics, Vol. 22 No. 4, pp. 180-186.
White, H. (1980), “A heteroskedasticity-consistent covariance matrix estimator and a direct test for
heteroskedasticity”, Econometrica: Journal of the Econometric Society, Vol. 48 No. 4, pp. 817-838.

Further reading
Kraus, A. and Litzenberger, R. (1983), “On the distributional conditions for a consumption-oriented
three moment CAPM”, The Journal of Finance, Vol. 38 No. 5, pp. 1381-1391.

Corresponding author
Sonal Babbar can be contacted at: researchatmydesk@gmail.com

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: permissions@emeraldinsight.com
Reproduced with permission of copyright
owner. Further reproduction prohibited
without permission.

S-ar putea să vă placă și