I. INTRODUCTION:
The paper has been divided into four sections. Section I presents
the overview. Section II presents review of International studies. Section III
presents review of Indian studies. Section IV concludes the paper.
II.
REVIEW OF INTERNATIONAL STUDIES:
Levy and Lerman (1988) have tested the
forecasting ability of portfolio creation strategies. They have analysed stochastic dominance with riskless asset (SDR)
rules and mean-variance with riskless asset (MVR). Standard stochastic
dominance (SD) rules and the traditional mean-variance (MV) and geometric mean
(GM) rules further supplement the results obtained through SDR and MVR.
Decision rules tested have been classified into three categories: (1) Investor
preference and distribution free SD rules; (2) Normally distributed MV rule and
(3) Logarithmic utility GM rule. Strategies tested include: (1) Buy and hold;
(2) Annual portfolio revision: terminal wealth maximization; and (3) One year
investment decision: expected utility maximization. The hypothesis that such
investment rules give returns superior to those attained by naive portfolio
strategies has been tested. Scenario analysis has been done by creating
portfolio having only risky assets, risky assets and riskless assets, lending
and borrowing portfolios, portfolio revision with or without transaction costs,
and varying investment holding periods and investment date. From the results so
obtained comments have been made on the efficiency (semi-strong form) of
capital markets. It was found that even after accounting for transaction costs,
gains superior to those of random portfolios can be made from ex-post data
using the MVR rule and the Second Degree Stochastic Dominance (SSDR) and Third Degree Stochastic Dominance (TSDR)
rule, provided riskless borrowing and lending is permitted. MVR, SSDR and TSDR also give higher terminal
wealth and ex-ante expected utility. First Degree Stochastic Dominance (FSDR)
rule creates very large portfolios with very high returns. MVR (relevant for
normal distribution only) gave similar ex-ante performance as given by SSDR and
TSDR for all the distributions. The GM rule portfolios also outperformed naive
portfolios. The findings indicate the presence of market inefficiency in the
weak and semi-strong form.
Ang and Bekaert
(2007) examined the relevance of dividend yields for predicting ex ante
superior returns on equities, cash flows and on interest rates. The predictive
power of dividend yields was observed only for short time periods and for short
term interest rates and almost of no significance in the long run. Also, growth
in dividends cannot be predicted from dividend yields. Causes for variation in
the dividend yields included discount rate and short term interest rates.
Earnings yield was found to be a significant factor in predicting future cash
flows. For predicting excess returns, short term interest rate is a significant
factor over short periods. Their analysis is based on data from United States
(US), United Kingdom (UK), France and Germany. For US, SandP
Data from June 1935-Decemeber 2001, for UK, FT Data from June 1953-December
2001, for Germany, DAX Data, June 1953-December 2001 and Morgan Stanley Capital
International (MSCI) data for US, UK, France and Germany from February
1975-December 2001 has been used. They had
build a nonlinear present value model with stochastic discount rates, short
rates and dividend growth to check the fit of regression based expected returns
with true expected returns. Their
research has a direct implication for future asset allocation studies which
will include dividends.
Griffin, Nardari and Stulz
(2007) investigated the trading and return data (weekly returns) for 46
countries to find out whether past high returns result in high turnover[i]. A large number of equities
markets exhibited a positive relation between the two variables. They have used
trivariate vector autoregression
(VAR) of market return, market volatility and turnover with weekly data from
1993 to 2003. Data has been collected from Datastream
International. It was found that a return shock was followed by significant
increase in turnover after 10 weeks in 24 countries. The relationship between
return and volume is stronger for developing countries as compared to developed
nations. The relationship between volume and return is very strong for
individual investors. High returns in the past resulted in high liquidity.
Return-volume relation is strong in countries with high corruption, high market
volatility, low correlation with world market and short sale constraints. A possible
explanation offered is that these countries have informational inefficient
capital markets and hence past returns carry informational value. Hence, many
other factors other than momentum trading affect return-volume relationship.
The authors have presented limitations of their theoretical model and empirical
analysis on return-volume relationship. Also, they did not include in their
analysis of the issue that does more volume imply more returns.
Liu (2007) discussed portfolio selection in stochastic
environments. Solution to dynamic portfolio choice problem for assets with
quadratic returns and constant relative risk aversion (CRRA) coefficient upto the stage of ordinary differential equation was
obtained. Three new applications
including stochastic equity return volatility as of Heston
model (1993) for equity portfolio have been solved. Stochastic variation in
investments was discussed in detail by Merton (1971) whereby the weights were
found as a solution to a nonlinear partial differential equation. Hence, only
approximate solutions can be obtained using nonlinear partial differential
equation. For explicit solutions to dynamic portfolio choice problems one may
use the methodology developed by Liu. The results so obtained are different
from static portfolio weights in the sense that negative dynamic portfolio
weights can be obtained with strictly positive risk premium and weights may not
be decreasing in risk aversion. The main contribution of this paper is that it
provides the theoretical framework for assigning weights in stochastic
volatility environment and explaining the rebalancing of portfolios over time.
A model in which investors have multiple priors (confidence
interval around the expected return) and aversion to ambiguity (minimization of
variance around the priors) was developed by Garlappi,
Uppal and Wang (2007). The model offers the benefit
of flexibility of degree of uncertainty that can be incorporated and gives a
close form expressions of optimal portfolio. The single prior of Bayesian
decision maker has been extended to consider investors with multiple priors. In
other words, it involves estimating expected returns with errors. This has been
done by an additional constraints of an upper and lower bound within which the
estimated value of expected return can fluctuate and minimization of this
confidence interval. It was found that portfolio weights using the multi-prior
approach were more balanced and required less rebalancing. For the empirical
analysis they have used data from Morgan Stanley Capital International (MSCI)
of monthly returns for Canada, France, Germany, Italy, Japan, Switzerland, the
United Kingdom and the United States from January 1970 to July 2001. Alternate
portfolios using different portfolio strategies including mean-variance,
minimum-variance, Bayes-stein and Ambiguity-averse
approach have been created. Higher out of sample Sharpe ratio was achieved for
ambiguity-averse portfolios as compared to classical and Bayesian models. The major limitation pointed by the authors
is that the model is not capable to produce new results in response to new
observations as it involves employing a set of probabilities.
Life-cycle portfolio choice with additive habit formation
preferences and uninsurable labour income risk was investigated
by Polkovnichenko (2007). For this habit-wealth
feasibility constraints have been derived which shows dependence on the future
scenario of worst possible income and habit. The model recommends conservative
portfolios in case future predicts income shocks. This has been done to ensure
that conservative portfolio can sustain the existing habits. The share of
equity portfolio increases with increasing wealth in the model as there will be
surplus available in excess of consumption and contingencies. Interestingly,
between moderate to high wealth, the share of equity portfolio in the total
wealth decreases and in the range from low to medium wealth, the equity
portfolio in total portfolio increases. The research problem is directed
towards understanding the portfolio allocation behaviour
of individuals over their life time. A life cycle model in general focuses on
the factors that influence individual portfolio allocation and changes in it
with age. This choice between saving and investment in risky asset has much
wider implications on portfolio selection theories, asset pricing theories,
microeconomics, economic development and public finance. The equity premium
puzzle still remains largely un-deciphered making the investment decision even
more complex. Parameterized standard life–cycle portfolio selection models in
general predict holding large proportion of equity portfolio by young
individual investors which may or may not be supported by empirical data. Using
empirical data it was shown that young investors hold more conservative
portfolios than middle-aged investors as also implied by their additive habit
model. The logic for this behaviour is that young
investors still need to accumulate saving to insure their habits. Similar
empirical results were also obtained by Heaton and Lucas (2000) and Faig and Shum (2002). Those investors with the objective of
bequest tend to accumulate more wealth than others. The model has been tested
for robustness by considering two extensions, involving borrowing against
future income and the other of flexibility in labour
supply. The contribution of the paper is in the manner it combines literature
on portfolio selection with labour income
uncertainty, finite horizon and additive habit formation preferences.
Lucas and Siegmann (2008) in their
empirical research on Hedge funds found that variance is an inadequate measure
of risk when there exists large downside risks. High Sharpe ratio will be
accompanied by high downside risk in the portfolio so constructed. “Expected Shortfall[ii]” has been tested an appropriate
alternate measure of risk. However,
empirical results pointed out the existing pitfalls of using expected shortfall
as the measure of risk. Optimal portfolios created on the basis of expected
shortfall may have skewness and kurtosis properties
inferior to mean-variance efficient portfolios. Quadratic shortfall measure
(MQSF) was found to be more promising than variance. It penalizes shortfalls more than linearly.
The two main contributions of the paper include finding payoff distributions
with modest shortfall and modelling optimal
portfolios for mean shortfall investors having equities and options in the
portfolio. Quadratic penalty on the
shortfall has been imposed for testing the robustness of the model. Most desirable
skewness and kurtosis characteristics were obtained
for quadratic shortfall as compared to variance and expected shortfall as
measure of risk. HFR data base has been used from January 1994 to December 2004
for hedge fund returns. Data had excess kurtosis and negative skewness. Static one month time horizon investor has been
considered for empirical analysis. Asset classes include stocks of SandP500, Saloman Brothers government bond index (SBWGU), a risk free
asset and hedge fund style indices. No short sale was permitted. It was shown
that the efficiency of expected shortfall criteria is dependent on the precise
shape of the extreme left hand tail of the asset return distribution. Inclusion
of data of period of crashes may result in selecting equities which may crash
in future. 13200 simulations were run for the 11 year sample data. Simulations
confirmed that expected shortfall performs does not perform better than mean
variance in case the return distributions are left skewed and fat tailed. The research
has contributed in pointing the need for an appropriate risk measure for
portfolio optimisation.
Brown and Sim (2009) introduced satisficing measures for evaluating financial positions in
terms of their ability to achieve target aspiration levels. It is also more
realistic to have either alternate benchmarks or fixed targets as compared to
having risk tolerance parameters. Using satisficing
measures, diversification has been rewarded. However, results for robustness
guarantees for such class of satisficing measures
remain ambiguous. The paper provided an axiomatical definition of the concept of satisficing measure. Representation theorem has been proved
whereby satisficing measures can be written as a
parametric family of risk measures. Since, satisficing
itself does not ensure diversification. Therefore, quasiconcavity
on satisficing measures has been imposed to ensure
diversification. Further, scale invariance has been imposed on the satisficing measure. While discussing portfolio optimisation using coherent satisficing
measures (CSM), an investor will choose a combination of risk free asset and
that risky asset which maximises the sharpe ratio. In the e-companion to the paper they have
discussed the computational example of creating a portfolio from a sample of 24
equities from SandP 500 using daily return from
January, 2004 until December, 2006. To have the probability of portfolio
returns higher than SandP Index they have developed a
Mixed Integer Programming (MIP) problem which in general is computationally
intractable but has been solved to most optimal solution by imposing a time
limit of two hours. A portfolio that maximises the
Conditional Value at Risk (CVaR) satisficing
measure has been described as desirable. Conditional Value at Risk (CVaR) satisficing measure was
found to be superior to using regular CVaR measure.
Saleh (2010) using data from Amman
Stock Exchange (ASE, Jordan) investigates the relevance of value investing for
the period 1980-2000. Value investing has been defined as investing in those
equities whose current market price is lower than some measure of fundamental
value. The paper tests the relevance of book-to-market equity and size in
explaining cross-sectional returns on equities. In emerging capital markets
like ASE, stock volatility was found to give suitable explanation for value
premium. For investigating the research problem, equally and value weighted
returns for portfolios formed on book-to-market equity ratio, market capitalisation and both of the previous mentioned factors
were simulated by the authors. Value - Glamour investing strategy was not found
to be working on ASE and the main reason for this was the volatility present on
ASE. It was because of higher volatility of small and high book-to-market
equities, that they outperformed small and low book-to-market equities.
However, large and low book-to-market equities gave higher returns than large
and high book-to-market equities on account of higher volatility of the latter.
Hence, volatility is an important variable to be modelled
in the existing framework of the Fama and French
(1993) Three Factor Model. At a significance level of less than 10 % it was
found that high volatility equities deliver higher returns as compared to low
volatility equities, when volatility is measured for one year period. During
boom period volatility has a significant effect on low book-to-market large
stocks. In unfavourable market conditions, volatility
has a significant effect on small and high book-to-market equities and large
and low book-to-market equities.
Wachter and Yogo
(2010) developed life-cycle consumption and portfolio choice model for
households having non-homothetic utility for luxury and basic goods. The model
has been successful in explaining the extent of growth of investments in risky
assets in the cross-section of households from the survey of consumer finances.
The authors have made four predictions which have been tested in the paper.
Censored regression model has been estimated for computing the expenditure share
for each category of nondurable goods and services. The sample consisted of stockholders from
1982-2003 consumer expenditure survey (data collected by the bureau of labour statistics through interviews representing 7,000
households). Data related to stockholders in the 1989-2004 survey of consumer
finances (data collected by Federal Reserve System using dual-frame sample
design representing 3,000 households) have been used for estimating a censored
regression model for the portfolio share. Implications of their model on the
portfolio choice have also been analysed. While analysing portfolio share by wealth it was observed that
portfolio share fall in wealth for households aged 26-45 and is flat or rising
for older households. In the non-homothetic model, there is a direct effect of
permanent income on portfolio choice. While analysing
portfolio share by age it was found that in nonhomothetic
life cycle model, the age effect of becoming risk averse is offset with the
increase in permanent income. Unemployment risk was not found to be having a
significant effect on the portfolio choice of most households. By unearthing
various dimensions of heterogeneity in risk aversion, their research is also
helpful in solving the existing puzzles related to benchmark asset pricing
theories. Heterogeneity of identical non-homothetic preference households has
been made integral part of their model subject to the idiosyncratic income
shocks over the life cycle. Also, risk aversion changes over time for
individual households in their model.
Non-homothetic life cycle model proposed by the authors is more
consistent in describing the sample than existing homothetic utility life-cycle
consumption and portfolio choice model.
III.
REVIEW OF INDIAN STUDIES:
Mishra (2001) examined the ability of
mutual funds as regards selectivity and timing skills and causes of
non-stationary beta of mutual funds. Sharpe, Treynor
and Jensen’s performance measure has been used for evaluating mutual fund
schemes. To evaluate selectivity and timing skills of mutual funds and beta
instability two steps generalized varying parameter estimation procedure has
been used. A sample of twenty four mutual fund schemes have been studied for
the period January 1992 to December 1996. Data was collected from economic times
mutual funds score board, BSE official directory, RBI’s monthly and annual
reports and reports on currency and finance. Out of the 24 schemes studied,
only four provided positive average return indicating poor performance of
mutual funds. It was also found that about eleven of the schemes were in the
high risk and low return quadrant. Diversification achieved by these schemes
was also low. Only two schemes had positive average Sharpe and Treynor index. Six mutual funds had a positive alpha and
even the alpha was found to be insignificant. The data analysis depicted a
dismal picture of the mutual fund performance, as most of the schemes were not
able to even offer risk free return. Eight schemes were found to display a net
positive selectivity, benefits of which were eroded by negative diversification
and high risk. A need for improvement in fund management was recommended. Beta
was found to be stable for eleven schemes. Hence, Jensen’s measure is a
suitable measure of portfolio performance for them. For remaining 13 schemes,
unstable beta (where beta is not fixed or changes because of portfolio
rebalancing or other random systematic factors) was observed. Generalized
varying parameter (GVP) estimates has been used in place of ordinary least
square estimates. This revealed positive timing skills for six schemes and
negative timing skills for seven schemes. A need for improvement in security
selectivity ability and timing skills was suggested in the thesis.
Manjunatha, Mallikarjunappa
and Begum (2006) tested the CAPM for Indian capital markets. 30 companies part
of BSE Sensex were included in the sample under
study. CMIE database and BSE website were used for collecting daily data for
the period January 3, 2000 to December 31, 2003. 28 portfolios are created
giving equal weightage to 5 equities at a time. Each
portfolio had 3 low beta equities and 2 high beta equities. Expected portfolio
return is calculated using ex post data for the period January 1, 2004 to
February 19, 2004. Contradicting to what CAPM purports, alpha was more than the
risk free rate. Slope of the regression equation for portfolio was found to be
negative and not equal to the risk premium. Inverse relationship between beta
and portfolio returns was observed.
Hence, using CAPM for creating portfolios for short periods was not
recommended.
Sudhakar and Kumar (2010) have presented
perceptions of the 500 investors investing in UTI Mutual Funds in Hyderabad.
Investment objective, risk tolerance, expected return on investment,
attractiveness of various mutual fund schemes and future prospects as perceived
by the investors has been analysed. They have used
Chi Square (
) for testing the independence of variables under study at 5
per cent level of significance. It was found that income and preference for
capital gain, dividend gain or both were statistically independent. However,
investor profile in terms of risk–return trade off, expectation of future
scenario and income were found to be statistically dependent. Their study is
useful in understanding the psychology of mutual fund investors.
Kumar (2010) studied investor preference for derivative and cash
market. Lack of awareness of derivatives resulted in most investors preferring
to invest in cash market. Also, liquidity, low investment and capital
appreciation favour investment in the cash market.
Study is based on a sample of 100 respondents availing brokerage services from
JRG Securities Limited (Erode). Chi-square test, ANOVA, paired t – test has
been used for the purpose of analysis. Using Chi square test, it was proved
that no relationship exists between monthly income and time period (short term,
medium term or long term) of investment. Using Chi square test it was proved
that no relationship exists between occupation of the investor (business,
professionals, service or housewife) and investment decision catalyst
(friends/relatives, agents, advertisements and others). Using t test, it was
proved that there exists a relationship between age of the investor and use of
margin funding in share trading.
Mehta and Chander (2010) tested Fama and French Three Factor Model on securities part of
BSE 500. Fama and French Model were found to have
significant explanatory power as regards cross-section of returns. Six
portfolios have been created to test the predictive power of the three factor
model. For determining the right investment strategy, the calendar effect has
also been examined. Monthly data from February 1999 to December 2007 from CMIE
Prowess for 219 companies for prices, BP ratio and size factor has been used.
Non parametric Krushal Wallis H Test and one
Parametric test (t-test and F-test) have been used to test the statistical
significance of difference of mean returns of monthly returns of various
portfolios constructed. Small size portfolios performed better than large size
portfolios along with higher volatility. Beta was still found to have the
maximum explanatory power for all the six portfolios. Fama
and French Three factor model explained more than 85% variation in four
portfolios and 80% variation in the remaining two portfolios. No evidence for
January or April effect was observed. However, November and December effect was
observed enabling investor to use Fama French model
for making portfolios giving superior returns.
Jeyachitra, Selvam
and Gayathri (2010) undertook an empirical study to
uncover portfolio risk and return relationship for securities from NSE Nifty.
Data of daily, weekly and monthly adjusted opening and closing share prices
from 1/4/2004 to 31/3/2009 for 40 actively traded securities part of SandP CNX Nifty Index has been collected from CMIE Prowess
for the analysis. Eight portfolios have been created in ascending order of beta
with five securities in each portfolio. A linear and positive relationship
between portfolio beta and return was observed. Hence, portfolios with high
beta gave higher returns. Exposure to unsystematic risk was reduced in the long
run (monthly) in portfolios with high beta. Holding high beta portfolios for a
month gave higher returns as compared to holding period of a week. In other
words, relationship between beta and monthly returns is more positive than
daily or weekly portfolios expected return and beta.
Vij and Tamimi
(2010) analysed the trade-off of risk and return for
sixty equities belonging to pharmaceutical industry listed on Bombay Stock
Exchange using CAPM Model. The data used in the study is from 2001 to 2007.
Regression analysis, t-test (5% significance level) and z-test have been used
in the research methodology. CAPM was found to be a good indicator exhibiting
linear and proportional trade-off between risk and return for the
pharmaceutical industry. The research objectives focussed
on analysing the effect of diversification and the
ability of return predictability of beta. Data from 2001 to 2007 for monthly
prices and index values has been taken from CMIE Prowess. BSE Sensex was taken as market proxy. 10 portfolios have been
created with 6 equities in each portfolio after arranging all the 60 securities
in ascending order of beta. Diversification in portfolios is achieved by
including equities having a particular range of beta. Alpha of the equities and
the portfolios was found to be equal to the risk free rate of return. A
correlation of 0.48 (for equities) and 0.78 (for portfolios) was observed
between beta and expected return indicating a positive relationship between the
two variables. Regular income investors are recommended to invest in low beta
portfolios. Speculators and capital gain seeking investors are recommended to
invest in high beta portfolios.
Banerjee (2011) investigated the price
performance of IPOs listed on National Stock Exchange (NSE) during the period
2001-2007. Long run (12 and 24 months) and short run (1 week and 6 months)
price performance has been analysed. 100 IPOs were
studied for the research problem. Short run and long run return and Wealth
Relative Index has been used as methodology for the analysis. Under-pricing as
a short run phenomenon was observed for equities listed on NSE. In the long
run, it was observed that corrections in prices take place and the market price
reaches the fair price. Further, portfolios of IPOs have been constructed based
on factors like issue size and age. When the age (starting from the date of
incorporation to the date of listing) is between 25-35 years, then the returns
generated are highest. Issue size greater than 100 lakhs
shares and less than equal to 230 lakhs shares tend
to give highest annualized raw returns.