Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
THE UNIVERSITY OF BURDWAN
Doctoral Program:
Business Administration
Doctoral Thesis
SELECTION OF PORTFOLIO WITH COMPARATIVE
ANALYSIS FOR SOME LISTED COMPANIES IN NIFTY
SOMA PANJA
Thesis Advisor: Professor Dilip Roy
Department of Business Administration
October 2012
Dedicated to my Parents and my Beloved Husband
\? fÉÅt ctÇ}t? wÉ {xÜxuç wxvÄtÜx à{tà à{x ãÉÜ~ ÜxÑÉÜàxw |Ç à{x à{xá|á
xÇà|àÄxw ÂfxÄxvà|ÉÇ Éy cÉÜàyÉÄ|É ã|à{ VÉÅÑtÜtà|äx TÇtÄçá|á yÉÜ fÉÅx
_|áàxw VÉÅÑtÇ|xá |Ç a|yàçÊ ã{|v{ |á ux|Çz áâuÅ|ààxw àÉ g{x hÇ|äxÜá|àç Éy
UâÜwãtÇ? UâÜwãtÇ? jxáà UxÇztÄ? \Çw|t |Ç ÑtÜà|tÄ yâÄy|ÄÄÅxÇà yÉÜ à{x
wxzÜxx Éy à{x WÉvàÉÜ Éy c{|ÄÉáÉÑ{ç |Ç TÜàá ;WxÑtÜàÅxÇà Éy Uâá|Çxáá
TwÅ|Ç|áàÜtà|ÉÇ<? |á t yt|à{yâÄ tÇw uÉÇt@y|wx ÜxáxtÜv{ ãÉÜ~ tÇw {tá ÇÉà uxxÇ
áâuÅ|ààxw xtÜÄ|xÜ àÉ tÇç Éà{xÜ hÇ|äxÜá|àç ÉÜ \Çáà|àâà|ÉÇ yÉÜ tÇç Éà{xÜ wxzÜxxA
WtàxM
;fÉÅt ctÇ}t<
i
Acknowledgements
First of all, I would like to express my deepest sense of gratitude to my guide
Professor Dilip Roy, retired Professor, Department of Business
Administration, The University of Burdwan for giving the opportunity of
developing my Ph.D. thesis under his supervision. I am grateful for his
scholarly guidance, inspiring and valuable suggestions, tirelessly monitoring
of the work and for his timely corrections with positive criticism, without which
this study would not have been completed or even started. I thank him for
believing in me, sometimes more than what I believe myself.
I am thankful to the authorities of The University of Burdwan for giving me
permission to do this research work.
I express my gratitude to all the members of the Centre for Management
Studies for creating a pleasant place to work and for providing me with all the
resources necessary to carry out this work with important suggestion.
I would like to thank to all my teachers, colleagues, co-researcher and friend
for their help and cooperation.
Finally, I would like to thank my family members, especially to my mother,
father and my husband Subroto and my son Suyash along with my in laws
and my elder sister and her family for being so supportive and
accommodative throughout the course of the study. They always supported
me and motivated and encouraged me to continue towards the completion of
this thesis. This achievement is theirs.
Date:
(Soma Panja)
ii
Table of Contents Contents Page
No:
Acknowledgements i
Table of Contents ii
Index of Tables V
Index of Figures Viii
List of Works X
List of Presentation xi
Notations xii
Chapter 1: Introduction
1.1 An Invitation to the Problem 1
1.2 Portfolio Management 2
1.3 Concept of Risk and Return 9
1.3.1 Concept of Risk 9
1.3.2 Concept of Return 11
1.4 Existing Approaches 13
1.4.1 Mathematical Programming Approach 13
1.4.2 Model Approach 20
1.5 Research gap and Research Problem 24
1.6 Preview of the Work 30
1.7 Scope of the Work 33
Chapter 2: Research Methodology
2.1 Introduction 36
2.2 Research Questions 36
2.3 Objectives 37
2.4 Research Design 38
2.4.1 Sampling Design 38
2.4.2 Statistical Design 39
2.5 Methods to be Used 59
iii
2.5.1 Optimization Method 59
2.5.2 Heuristic Method 60
2.5.3 Model Based 62
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
3.1 Introduction 65
3.2 Earlier works 66
3.3 Optimization Framework 70
3.4 Heuristic Framework 72
3.5 Closeness between Heuristic and Optimization Framework 74
3.6 Data Analysis and Interpretation 74
3.7 Conclusion 99
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
4.1 Introduction 100
4.2 Earlier works 101
4.3 Optimization Framework 105
4.4 Heuristic Framework 106
4.5 Closeness between Heuristic and Optimization Framework 108
4.6 Data Analysis and Interpretation 109
4.7 Conclusion 130
Chapter 5: Value Based Selection of Portfolio: A Heuristic Approach
5.1 Introduction 131
5.2 Earlier works 132
5.3 Choice of weights under different value systems 134
5.4 Data Analysis and Interpretation 136
5.5 Conclusion 145
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under Model Approach
6.1 Introduction 147
6.2 Earlier works 148
6.3 Sharpe’s Single Index Model 152
iv
6.4 Sharpe’s cut-off Principle 152
6.5 Optimum framework under Value System Sharpe’s Single Index Model
154
6.6 A Near Optimum Approach 155
6.7 Comparative study 156
6.8 Conclusion 164
Chapter 7: End Remarks
7.1 Importance of value system 165
7.2 Closeness between Optimum Solution and Heuristic Solution
167
7.3 Limitations of the Current work 170
7.4 Future Scope 171
References 173
v
Index of Tables Table No:
Details Page No:
3.1 Weights of optimum portfolio and heuristic portfolios under -0.0057 minimum return constraint
76
3.2 Expected return and risk of three different portfolios under – 0.0057 as the minimum return constraint
77
3.3 Weights of optimum portfolio and heuristic portfolios under – 0.0047 as the minimum return constraint
79
3.4 Expected return and risk of three different portfolios under – 0.0047 as the minimum return constraint
80
3.5 Weights of optimum portfolio and heuristic portfolios under – 0.0037 as the minimum return constraint
82
3.6 Expected return and risk of three different portfolios under – 0.0037 as the minimum return constraint
83
3.7 Weights of optimum portfolio and heuristic portfolios under – 0.0027 as the minimum return constraint
84
3.8 Expected return and risk of three different portfolios under – 0.0027 as the minimum return constraint
85
3.9 Weights of optimum portfolio and heuristic portfolios under -0.0017 as the minimum return constraint
87
3.10 Expected return and risk of three different portfolios under – 0.0017 as the minimum return constraint
88
3.11 Weights of optimum portfolio and heuristic portfolios under -0.0007 as the minimum return constraint
89
3.12 Expected return and risk of three different portfolios under – 0.0007 as the minimum return constraint
90
3.13 Weights of optimum portfolio and heuristic portfolios under 0.0003 as the minimum return constraint
92
3.14 Expected return and risk of three different portfolios under 0.0003 as the minimum return constraint
93
3.15 Weights of optimum portfolio and heuristic portfolios under 0.00065 as the minimum return constraint
94
3.16 Expected return and risk of three different portfolios under 0.00065 as the minimum return constraint
95
vi
3.17 Euclidean Distance between pessimistic and optimum solutions and between optimistic and optimum solutions
97
4.1 Optimum weights of fifteen securities under different values of coefficient of optimism
109
4.2 Weights of heuristic optimistic, heuristic pessimistic, heuristic risk planner and heuristic random selector solutions
111
4.3 Expected return and risk of the optimum portfolio under different α values
112
4.4 Values of expected returns and risks of heuristic portfolios
113
4.5 City Block Distance between the optimistic & optimum choices of portfolios and pessimistic & optimum choices of portfolios
114
4.6 City Block Distance between the optimistic & optimum choices of portfolios and risk planner & optimum choices of portfolios
116
4.7 City Block Distance between the optimistic & optimum choices of portfolios and random selector & optimum choices of portfolios
118
4.8 City Block Distance between the pessimistic & optimum choices of portfolios and risk planner & optimum choices of portfolios
120
4.9 City Block Distance between the pessimistic & optimum choices of portfolios and random selector & optimum choices of portfolios
122
4.10 City Block Distance between the pessimistic & optimum choices of portfolios and random selector & optimum choices of portfolios
124
4.11 City Block Distance between optimistic and optimum solutions, between pessimistic and optimum solutions, between risk planner and optimum solutions and between random selector and optimum solutions
127
5.1 Weights of the optimum portfolios of fifteen selected securities under different α coefficients
137
5.2 Expected return and risk of the fifteen selected 138
vii
securities under different α coefficients
5.3 Weights of two heuristic portfolios taken into consideration
139
5.4 Weights of the heuristic portfolios of fifteen selected securities under different α coefficients
140
5.5 Expected return and risk of heuristic portfolios under different α coefficients
142
5.6 City Block Distances between optimum portfolios and heuristic portfolios under different α coefficients
143
6.1 Weights of the optimum portfolios under Sharpe’s Single Index Model of ten securities under different α values
157
6.2 Expected return and risk of the optimum portfolios under Sharpe’s Single Index Model under different values of α
158
6.3 Weights of the near optimum portfolios of ten securities under different α values
159
6.4 Expected return and risk of the near optimum portfolios under different values of α
160
6.5 CBD between the optimum portfolio under Sharpe’s Single Index Model and Sharpe’s cut-off principle portfolio
161
6.6 CBD between the optimum portfolios under Sharpe’s Single Index Model and the near optimum portfolios
162
viii
Index of Figures
Figure
No: Page No:
3.1 Expected return and risk under -0.0057 minimum return constraint
78
3.2 Expected return and risk under -0.0047 minimum return constraint
80
3.3 Expected return and risk under -0.0037 minimum return constraint
83
3.4 Expected return and risk under -0.0027 minimum return constraint
86
3.5 Expected return and risk under -0.0017 minimum return constraint
88
3.6 Expected return and risk under -0.0007 minimum return constraint
91
3.7 Expected return and risk under 0.0003 minimum return constraint
93
3.8 Expected return and risk under 0.00065 minimum return constraint
96
3.9 ED of Optimistic and Pessimistic Solution 98
4.1 City Block Distance between Optimum Portfolio and Optimistic Portfolio and Optimum Portfolio and Pessimistic Portfolio
115
4.2 City Block Distance between Optimum Portfolio and Optimistic Portfolio and Optimum Portfolio and Risk Planner's Portfolio
117
4.3 City Block Distance between Optimum Portfolio and optimistic Portfolio and Optimum Portfolio and Random Selector's Portfolio
119
4.4 City Block Distance between Optimum Portfolio and Pessimistic Portfolio and Optimum Portfolio and Risk Planner's Portfolio
121
ix
4.5 City Block Distance between Optimum Portfolio and Pessimistic Portfolio and Optimum Portfolio and Random selector Portfolio
123
4.6 City Block Distance between Optimum Portfolio and Risk Planner's Portfolio and Optimum Portfolio and Random Selector's Portfolio
125
4.7 City Block Distance between Optimum Portfolio and Heuristic Portfolios
128
5.1 CBD between Optimum and Heuristic Portfolios 144
6.1 City Block Distance between Sharpe’s Single Index Model and Cut-off principle and Sharpe’s Single Index Model and Near Optimum Portfolio
163
x
This thesis is based on the following published works and a few works
submitted for publication:
Published:
1. Roy, D & Panja,S. (2011): Heuristic selection of portfolio based
on coefficient of optimization, International Journal of Financial
Management, (ISSN: 2229-5682), Volume 1 Issue 1, 2011.
2. Roy, D, Mitra, G. & Panja,S. (2010): Closeness between Heuristic
and optimum selection of portfolio: An Empirical analysis, GITAM
Review of International Business, (ISSN No. 0974-357X), Vol 2,
Issue 2, January 2010.
Submitted for Publication
1. Panja, S. & Roy, D. (2011): A comparative study on Sharpe’s cut
off principle and a near optimum principle for portfolio selection
(Communicated for Publication to Journal of Financial and
Quantitative Analysis- Cambridge Journal).
2. Panja, S. & Roy, D. (2011): Risk based Selection of Portfolio:
Heuristic Approach (Communicated for Publication to The
Journal of Portfolio Management-Institutional Investor Journal).
xi
Conference Attended
Presented Paper titled “Closeness between Heuristic and optimum selection
of portfolio: An Empirical analysis” International conference on Advanced Data
Analysis, Business Analytics and Intelligence by IIM Ahmedabad (6-7 June
2009)
xii
Notations
ijR = Daily return of the ‘i’th security, where, i= 1,2,3,……………….,n. and
j= 1,2,3,………..,k.
σi = Standard Deviation of the ‘i’th security
i = Average return of the ‘i’th security
μmin = Minimum return of the security in the portfolio
μmax = Maximum return of the security in the portfolio
OPR = Expected return of the optimum portfolio
o
iw = Optimum weight of the ‘i’th security
Ri =Expected return of the ‘i’th security
n = Number of the securities in the portfolio
OP = Risk of the optimum portfolio
rij = Correlation coefficient between ‘i’th and ‘j’th securities
i = Standard Deviation of ‘i’th security.
HO
iw = Heuristic optimistic weight of the ‘i’th security
μi = Expected return of ‘i’th security
HOPR = Expected return of the optimistic portfolio
HOP = Standard Deviation of the heuristic optimistic portfolio.
HPiw = Heuristic pessimistic weight of ‘i’th security
HPPR = Expected return of the pessimistic portfolio
HPP = Standard Deviation of the pessimistic portfolio
xiii
α = Coefficient of optimism of the investor (0≤α≤1)
∑ = Dispersion matrix of the securities
~w = Weight vector
``R = Expected return vector
HRPiw = Heuristic risk planner weight of the ‘i’th security
HRPPR = Expected return of the heuristic risk planner portfolio
HRPP = Standard Deviation of the heuristic risk planner portfolio
HRSpR = Expected return of the heuristic random selector portfolio
HRSp = Standard Deviation of the heuristic random selector portfolio
Hw Weight of the ‘i’th security in the heuristic portfolio with coefficient of
optimism as α
HR Expected return of the general heuristic portfolio
H Standard deviation of the general heuristic portfolio
mR Rate of return on the market index, a random variable
i = Expected value of the component of return independent of the market’s
performance
i The expected change in the rate of return on stock ‘i’ associated with a
1% change in the market return
ie = Error component
xiv
2m Variance of market return
2ei = Variance of error term
iR Average return of ‘i’th security
pR Return of the portfolio
iw = Weight of ‘i’th security
p Standard deviation of the portfolio
2p = Variance of returns or portfolio risk
iC = Cut-off rate
*C = A candidate of iC
V{tÑàxÜ D
\ÇàÜÉwâvà|ÉÇ
Chapter 1: Introduction
1 | P a g e
1.1 An invitation to the problem
In financial management, risk management is an
important concept. Financial Risk Management can be defined
as an act of creating economic value of the firm by effectively
managing the exposure to the risk, with judicious use of several
financial instruments and sophisticated techniques. The different
type of exposure to risk, a firm is subjected to mainly involve
credit risk and market risk. Financial risk management can be
qualitative and quantitative in approach.
In order to determine the optimal asset allocation
strategies mathematical models are vastly used by the corporate
investors and portfolio managers respectively. The key
organizations in finance are households, business firms,
financial intermediaries and capital markets. The tradition in
neoclassical economics is to consider the existence of
households, their tastes and their endowments as exogenous to
the theory. But other economic organizations are regarded as
primary because of the functions they serve and are therefore
endogenous to the theory. Giving concentration more on
household, there are two players – consumer and investor. The
consumer chooses how much of her income and wealth to
allocate to current consumption and thereby, how to save for
Chapter 1: Introduction
2 | P a g e
future consumption. As investor, the household solves the
problem to determine the fractional allocation of her savings
among the available investment opportunities.
Risk management plays a critical role in determining the
financial dynamics of the corporate investors. However, risk
management holds equal significance when individual investors
and their investment objectives are considered. Therefore for
this purpose asset management is important for the individual
investors. If a person owns more than one asset for investment,
she has an investment portfolio. A portfolio consists of more
than one asset. The main aim of the portfolio owner is to
enhance the value of portfolio by selecting investments that yield
good returns.
1.2 Portfolio Management
Portfolio Management refers to the science of analyzing
the strengths, weaknesses, opportunities and threats for
performing wide range of activities related to the one’s portfolio
for maximizing the return at a given risk. It helps in making
selection of Debt Vs Equity, Growth Vs Safety, and various other
tradeoffs. Portfolio management involves the task of taking
decisions about investment policy and mix, matching
Chapter 1: Introduction
3 | P a g e
investments to the objectives of the investors, allocating assets
for individuals and institutions and balancing risk against
performance. Portfolio management has a great importance in
theory of finance. Managing a portfolio involves inherent risks.
Portfolio management is goal driven and target oriented.
Constructing a portfolio involves making wide range of decisions
regarding buying or selling of stocks, bonds, or other financial
instruments. In portfolio management both time and magnitude
are very important.
Portfolio optimization plays a critical role in determining
portfolio strategies for investors. In portfolio optimization
investors want to either maximize portfolio returns or minimize
portfolio risks. Since return is compensated based on risk,
investors have to balance the risk-return tradeoff for their
investments and it depends upon investors risk-return
preferences. So, a single optimized portfolio is not in a position
to satisfy all investors.
The traditional mean variance optimization approach fails
to meet the demand of investors who have multiple investment
objectives. In order to achieve the multiple objective of the
investor and to satisfy the aim of obtaining the optimal portfolio it
is necessary to understand the proportion of investment of
Chapter 1: Introduction
4 | P a g e
different assets in the portfolio. That means fixation of asset
weights correctly of a given portfolio is of prime importance from
investors point of view. The above discussed premises points at
the ultimate necessity of fixation of assets’ weight as the
decision variable in the portfolio optimization problems. The use
of portfolio weights for model generation may be especially
important in a setting where the expected return and the
variance are assumed to be correctly predicted and are
expected to show same behavior in future. The use of portfolio
weights to measure the performance of trading strategies was
pioneered by Cornell (1979). Cornell’s measure was modified
by Copeland and Mayers (1982) to analyze Value Line rankings.
The concept was applied to weight-based measure of mutual
fund performance (Grinblatt and Titman, 1993).
The weight-based performance measures of portfolio are
rather simple. For example, an investor if knows that the returns
of assets are likely to be higher or lower than expected by the
market, then, other things remaining same, the investor can
earn profit by changing his portfolio weights toward those assets
whose returns are likely to be higher than expected and away
from those assets whose returns are likely to be lower. In other
words the covariance between the change in a portfolio’s
Chapter 1: Introduction
5 | P a g e
weights and subsequent abnormal asset returns may be used to
measure performance of portfolio.
Most of the portfolio measurement techniques are return-
based, and involve regressing the return of a portfolio on some
benchmark return. The measure of performance is the intercept
in the regression. The minimal information requirement of the
return methodologies is cited as one of the strength. An investor
needs only returns on the managed portfolio and the benchmark
in portfolio optimization problem. The overemphasis on the risk
and return dimensions of the portfolio optimization often
overlooks the necessity of the potential important information
regarding the composition of the portfolio. Previously, portfolio
weights were used with unconditional moments to measure
portfolio performance. However, the inclusion of return-based
measures into a conditional framework changed the results
(Ferson and Schadt, 1996) and hence made it interesting to
consider formulation and modeling in weight-based measures of
performance.
The manner in which the choice between the different
courses of action or inaction is made plays a critical role in
deciding the effectiveness of a portfolio manager. The easiest
way to make the decision is to evaluate from the array of
Chapter 1: Introduction
6 | P a g e
predicted risk and return of the securities in the portfolio and
select the best combination. But the question which remains to
be addressed is the search and applicability of a common model
that can be used to capture the risk and returns, the different
prediction mechanisms, the role played by different information
dynamics which are supposed to influence the assessment of
the risks and returns, and so many prior expectations that may
be brought to influence the proportion of security in the design of
the optimum portfolio.
Since the middle of twentieth century, financial economist
or practitioner and statisticians had been measuring the
performance of a managed portfolio from various angles. Even
after years of research, several issues remain unsolved. The
classical mean variance model is aimed at satisfying the
optimizing needs of the risk averse investors. However, in real
world investors exhibit a multitude of risk profiles which explicitly
points to the fact and necessity of a model which can take into
account the dynamics of the different categories of investors on
the parameter of risk taking aptitude and attitude. Thus, it
becomes evident that the investors are influenced by their
nature or value system. According to Heller (1971) there are six
Chapter 1: Introduction
7 | P a g e
values of importance. Out of which, propensity to take risk is
most important value system (Jauch and Glueck, 1988).
Nature of the investors may not be same and their
investment needs depend on their nature. Some investors are
risk taker, some are risk aversive, some other investors invest
their wealth after a proper planning, who are known as risk
planners and the rests are random selectors who randomly
select the assets without giving any importance on expected
return and risk. Thus, if an investor can correctly predict the
proportion of amount to be invested in each assets of the
portfolio then right decision can be taken to earn maximum
utility.
An important unsolved issue is how to handle the
dynamic behavior of a managed portfolio. Not only because of
the existence of time-varying required returns in a portfolio but
also due to investor’s strategy or other influencing factors,
management of portfolios become difficult. Thus creation of
optimal portfolio strategy for investors of various risk profiles
becomes a lurking question in the financial risk management
scenario. An attempt has been made in this work to formulize
and propose a multi-objective approach to portfolio optimization
problems. In this work portfolios risk and return are optimized
Chapter 1: Introduction
8 | P a g e
and various portfolio optimization models are integrated.
Detailed analysis based on heuristic weight generation and
subsequent optimization and application of the model are
provided and compared to portfolio generated through the
mean-variance approach and Sharpe’s approach.
The basics and ideas of investment portfolio
management are used in various dimensions as discussed:
Application Portfolio Management: it involves
management of complete group or subset of computer software
applications in a portfolio. As the applications of software include
maintenance cost and development cost, these can be
considered as investment. The decisions regarding purchasing
new software or modifying existing software are important parts
in application portfolio management.
Product Portfolio Management: it means grouping of
major products developed or sold by businesses into (Logical)
portfolio. These products are arranged according to major line-
of- business or business segment. In this case investment
decision involves development of new product or modifying the
existing product or discontinuation any other product.
Chapter 1: Introduction
9 | P a g e
Project Portfolio management or initiative portfolio
management: it includes a specified beginning and end; precise
and limited collection of desired results or work products and
management team for taking the initiative and utilizing the
resources.
1.3 Concept of Risk and Return
1.3.1 Concept of Risk
The earliest definition of risk was given by Knight (1921).
According to him, risk is measurable uncertainty. Many authors
have given various definitions of risk. Risk is a concept that
denotes a potential negative impact to an asset. However, there
was an attempt made by Holton (2004) to summarize the most
relevant definition of risk to give a more general concept of risk.
According to him, risk is the exposure to a proposition of which
one is uncertain. A finance relevant definition given by Jorion
(2000) defines risk as the volatility of the expected results on the
value of assets and liabilities of interest. Portfolio management
relevant definition given by Cool (1999) defines risk as the
absolute value of probable loss.
From the economic point of view, risk is any event or
action that may adversely affect an organization's ability to
Chapter 1: Introduction
10 | P a g e
achieve its objectives and execute its strategies. In finance,
financial risk is essentially any risk associated with any form of
financing. The existence of risk means that the investor can only
associate a single number or payoff with investment in any asset
with certain probability associated with it. A common property of
investment opportunities is that their actual returns might differ
from what has been expected; or in short: they are risky. When
the actual return is lower than the expected outcome, it is known
as downside risk where as when the deviation from the actual
return is more than the expected outcome is known as upside
risk.
Risk reflects not only the dangers associated with an
investment, but also the chances. Therefore, a risky situation is
one in which surprises and unexpected developments might
occur. Volatility of return is this type of risk measure which is
one of the foundations of portfolio theory. Risk can be measured
in different ways; viz. semi-variance which measures only the
negative deviations from the expected value. More recently,
Value at Risk (VaR) has been used by the authors. It means the
maximum loss within a certain period of time with a given
probability. A third type of risk refers to a situation of danger or
peril. In finance, this concept is applied as an aspect of risk
Chapter 1: Introduction
11 | P a g e
measure as applied in circumstances where catastrophes ought
to be prevented; investors ought to be saved from hazardous
situations. However, in finance there are two types of risk-
systematic risk or un-diversifiable risk and unsystematic risk or
diversifiable risk. Systematic risk is the market risk or the risk
that cannot be diversified away. Systematic risk affects the
whole economy of a country. A perfectly diversifiable portfolio
also carries some systematic risk which we cannot avoid.
Unsystematic risk or diversifiable risk is the risk which is
associated with individual assets and it differs from asset to
asset. It is industry or company specific risk. Unsystematic risk
can be diversified away by including a number of assets in the
portfolio. However, unsystematic risks do not present enough
information about the overall risk of the entire portfolio. The
impact of risk has two components: uncertainty and exposure.
The chance of facing risk is uncertainty and exposure is the
amount of the possible loss if the risk has been faced.
1.3.2 Concept of Return
Return in its optimistic outlook can be defined as any
potential gain which is supposed to be over and above than the
amount invested in any asset. In other words, return is the ratio
of money gained or lost (whether realized or unrealized) on an
Chapter 1: Introduction
12 | P a g e
investment relative to the amount of money invested. A general
investor always tries to earn some return from any investment.
The return can be calculated over a single period, or expressed
as an average over multiple periods of time. Return can be
calculated in various ways viz. arithmetic return, geometric
return, logarithmic return. Return may be annual return and
annualized return. An annual return is a single-period return,
while an annualized return is a multi-period or arithmetic
average return.
Personalized investment returns are of recent origin and
are in much demand amongst the investor’s community. The
demand for personalized investment returns holds the argument
that the fund returns may not be the actual account returns
which are based upon the actual investment account transaction
history. This occurs because investments may have been made
on various dates and additional purchases and withdrawals may
have incurred on various dates and the related amount is
varying and thus is unique to the particular account. The fund
returns may be more or less than the account return. More and
more fund and brokerage firms have begun providing
personalized account returns on investor's account statements
in response to this need.
Chapter 1: Introduction
13 | P a g e
1.4 Existing Approaches
1.4.1 Mathematical Programming Approach
The selection of a best element (with regard to some
criteria) from some set of available alternatives in parlance of
mathematics, computer science, or management science is
known as mathematical
optimization (alternatively, optimization or mathematical
programming).In the simplest case, an optimization
problem consists of maximizing or minimizing a real function by
systematically choosing input values from within an allowed set
and computing the value of the function. The generalization of
optimization theory and techniques to other formulations
comprises a large area of applied mathematics. More generally,
optimization includes finding "best available" values of some
objective function given a defined domain, including a variety of
different types of objective functions and different types of
domains (Wikipedia,2012).
The following are applications of mathematical
optimization and include the following:
Convex programming studies the case when the
objective function is convex (minimization)
Chapter 1: Introduction
14 | P a g e
or concave (maximization) and the constraint set is convex. This
can be viewed as a particular case of nonlinear programming or
as generalization of linear or convex quadratic programming
(Wikipedia, 2012).
Linear programming (LP), a type of convex
programming, studies the case in which the objective
function f is linear and the set of constraints is specified using
only linear equalities and inequalities. Such a set is called
a polyhedron or a poly tope if it is bounded (Wikipedia, 2012).
Second order cone programming (SOCP) is a convex
program, and includes certain types of quadratic programs
(Wikipedia, 2012).
Semidefinite programming (SDP) is a subfield of
convex optimization where the underlying variables are semi
definite matrices. It is generalization of linear and convex
quadratic programming (Wikipedia, 2012).
Conic programming is a general form of convex
programming. LP, SOCP and SDP can all be viewed as conic
programs with the appropriate type of cone (Wikipedia, 2012).
Chapter 1: Introduction
15 | P a g e
Geometric programming is a technique whereby
objective and inequality constraints expressed
as polynomials and equality constraints as monomials can be
transformed into a convex program (Wikipedia, 2012).
Integer programming studies linear programs in which
some or all variables are constrained to take on integer values.
This is not convex, and in general much more difficult than
regular linear programming (Wikipedia, 2012).
Quadratic programming allows the objective function to
have quadratic terms, while the feasible set must be specified
with linear equalities and inequalities. For specific forms of the
quadratic term, this is a type of convex programming (Wikipedia,
2012).
Fractional programming studies optimization of ratios of
two nonlinear functions. The special class of concave fractional
programs can be transformed to a convex optimization problem
(Wikipedia, 2012).
Nonlinear programming studies the general case in
which the objective function or the constraints or both contain
nonlinear parts. This may or may not be a convex program. In
Chapter 1: Introduction
16 | P a g e
general, whether the program is convex affects the difficulty of
solving it (Wikipedia, 2012).
Stochastic programming studies the case in which
some of the constraints or parameters depend on random
variables (Wikipedia, 2012).
Robust programming is, like stochastic programming,
an attempt to capture uncertainty in the data underlying the
optimization problem. This is not done through the use of
random variables, but instead, the problem is solved taking into
account inaccuracies in the input data (Wikipedia, 2012).
Combinatorial optimization is concerned with problems
where the set of feasible solutions is discrete or can be reduced
to a discrete one (Wikipedia, 2012).
Infinite-dimensional optimization studies the case
when the set of feasible solutions is a subset of an infinite-
dimensional space, such as a space of functions (Wikipedia,
2012).
Heuristics and metaheuristics make few or no
assumptions about the problem being optimized. Usually,
heuristics do not guarantee that any optimal solution need be
Chapter 1: Introduction
17 | P a g e
found. On the other hand, heuristics are used to find
approximate solutions for many complicated optimization
problems (Wikipedia, 2012).
Constraint satisfaction studies the case in which the
objective function f is constant (this is used in artificial
intelligence, particularly in automated reasoning) (Wikipedia,
2012).
Dynamic programming studies the case in which the
optimization strategy is based on splitting the problem into
smaller sub-problems. The equation that describes the
relationship between these sub-problems is called the Bellman
equation (Wikipedia, 2012).
Mathematical programming with equilibrium
constraints is where the constraints include variational
inequalities or complementarities (Wikipedia, 2012).
A subset of mathematical programming approach, the
linear programming is a class of optimization problems. The
linear programming problems have one objective function and
the set of constraints with linear equalities and inequalities.
Because of the effectiveness and robustness of linear program
Chapter 1: Introduction
18 | P a g e
solving algorithms this techniques are useful for portfolio
rebalancing problems (Wikipedia, 2012).
Chekhlov et al (2004) stated that portfolio allocation
problems can efficiently be handled with linear programming
based algorithms. The techniques are attractive to the investors
as it demonstrates the problems with thousands of instruments
and scenarios. The portfolio mean-variance optimization
techniques are a class of quadratic programming problems. The
quadratic programming Optimization models can lead to non-
convex multi extrema problems.
Modern Portfolio theory: Markowitz, Sharpe, Tobin and
more
The modern portfolio theory introduced by Harry
Markowitz (1952) was based on linear programming problem. In
the model he defined the linear programming as either by
maximizing the return subject to certain amount of risk or
minimizing the risk subject to certain amount of return. He
proposed that investors should focus on selecting portfolios
based on their joint risk-reward feature. The expected return of
any portfolio can be measured by using the historical returns of
each asset on the portfolio. Various statistical measures such as
Chapter 1: Introduction
19 | P a g e
average (return), standard deviation and linear correlation are
used to measure the volatility of the portfolio. Markowitz (1959)
used volatility and expected return as proxies for risk and
reward. Markowitz defined an optimal way of selecting a
portfolio by balancing the risk and reward features of the
portfolio. According to him a rational investor should select a
portfolio that lies on the efficient frontier. Modern Portfolio
Theory as propounded by Harry Markowitz has introduced the
concept of efficient frontier. "Efficient Frontier" can be defined as
the combination of assets, i.e. a portfolio, if it has the best
possible expected level of return for its level of risk (usually
proxied by the standard deviation of the portfolio's return).
Markowitz work was expanded by James Tobin (1958) by
adding a risk-free asset to the analysis. He showed that by using
leverage or deleverage on the portfolios on the efficient frontier it
was possible to outperform them in terms of their risk and
reward relation. To prove it he introduced the concepts of
“Capital Market Line” and “super-efficient portfolio”.
In the Capital Asset Pricing Model (CAPM) Sharpe (1964)
pointed out that the market portfolio lies on the efficient frontier
and is also actually Tobin's super-efficient portfolio. CAPM first
introduced the concept of “beta” and relates an asset's expected
Chapter 1: Introduction
20 | P a g e
return to its beta. He showed that according to the risk apatite,
all investors should hold the market portfolio where it is
leveraged or de-leveraged with positions in the risk-free asset.
1.4.2 Model Approach
The description of a system using mathematical concepts
and language is known as a mathematical model. Mathematical
modeling is the process of developing a mathematical model. In
order to explain a system and to study the effects of different
components, and also to make predictions about behavior
mathematical models comes handy. Mathematical models can
take many forms, including but not limited to dynamical
systems, statistical models, differential equations, or game
theoretic models. A mathematical model usually describes a
system by a set of variables and a set of equations that
establish relationships among the variables. There are six basic
groups of variables namely: decision variables, input variables,
state variables, exogenous variables, random variables, and
output variables. Since there can be many variables of each
type, the variables are generally represented by vectors.
Decision variables are sometimes known as independent
variables. Exogenous variables are sometimes known as
parameters or constants. The variables are not independent of
Chapter 1: Introduction
21 | P a g e
each other as the state variables are dependent on the decision,
input, random, and exogenous variables. Furthermore, the
output variables are dependent on the state of the system
(represented by the state variables).
Objectives and constraints of the system and its users
can be represented as functions of the output variables or state
variables. The objective functions will depend on the perspective
of the model's user. Depending on the context, an objective
function is also known as an index of performance, as it is some
measure of interest to the user. Although there is no limit to the
number of objective functions and constraints a model can have,
using or optimizing the model becomes more involved
(computationally) as the number increases (Wikipedia, 2012).
Classifying mathematical models
Many mathematical models can be classified in some of
the following ways:
Linear vs. nonlinear: Mathematical models are usually
composed by variables, which are abstractions of quantities of
interest in the described systems, and operators that act on
these variables, which can be algebraic operators, functions,
differential operators, etc. If all the operators in a mathematical
Chapter 1: Introduction
22 | P a g e
model exhibit linearity, the resulting mathematical model is
defined as linear. A model is considered to be nonlinear
otherwise (Wikipedia, 2012).
Deterministic vs. probabilistic (stochastic):
A deterministic model is one in which every set of
variable states is uniquely determined by parameters in the
model and by sets of previous states of these variables.
Therefore, deterministic models perform the same way for a
given set of initial conditions. Conversely, in a stochastic model,
randomness is present, and variable states are not described by
unique values, but rather by probability distributions (Wikipedia,
2012).
Static vs. dynamic: A static model does not account for
the element of time, while a dynamic model does. Dynamic
models typically are represented with difference equations or
differential equations (Wikipedia, 2012).
Discrete vs. Continuous: A discrete model does not
take into account the function of time and usually uses time-
advance methods, while a Continuous model does. Continuous
models typically are represented with f (t) and the changes are
reflected over continuous time intervals (Wikipedia, 2012).
Chapter 1: Introduction
23 | P a g e
Deductive, inductive, or floating: A deductive model is
a logical structure based on a theory. An inductive model arises
from empirical findings and generalization from them. The
floating model rests on neither theory nor observation, but is
merely the invocation of expected structure. Application of
mathematics in social sciences outside of economics has been
criticized for unfounded models. Application of catastrophe
theory in science has been characterized as a floating model
(Wikipedia, 2012).
In portfolio management a lot of the research has been
done in modeling the uncertainty of the value of assets on a
portfolio and the relations between them, often heavily relying in
probability theory and statistics. These models are very often
used to simulate possible future scenarios through extensive
computer programs. These models may play a vital role in the
decision making process of the investors and they represent
solely another tool of analysis. Although the use of a model
constitutes another risk by itself, it may enable portfolio and risk
managers to explicitly take into consideration some of the
uncertainty they face and to quantify and estimate as accurately
as possible the risks they take (Wikipedia, 2012).
Chapter 1: Introduction
24 | P a g e
1.5 Research gap and Research Problem
Markowitz (1952) is credited to proclaim the Numero Uno
position in successfully quantifying the two basic conflicting
objectives of investing in a portfolio viz. maximizing expected
return and minimizing risk. Since the formulation of the modern
portfolio theory, his work has attracted the attention of the
academic world and has been instrumental in providing the
fundamental direction to address the issue of portfolio
optimization. Unfortunately, in the real world of investment
management, the Markowitz framework has had surprisingly
little impact. The reasons are, first, investors tend to focus on
small segments of their potential investment universe. They
select the undervalued assets and finds assets with positive
momentum, or identifying relative value trades. Unfortunately,
the Markowitz model needs expected returns to be specified for
every component of the relevant universe which is unrealistic.
But in practice this is typically defined by a broad benchmark.
Secondly, the investors put emphasis on the weights in a
portfolio. They are not much involved in balancing the expected
returns against the contribution to portfolio risk which is the
relevant margin in the Markowitz framework. The application of
Markowitz model to ascertain the weights of the assets to
construct the portfolio often results in extreme values which
seems to be computationally justified but lacks logical and
Chapter 1: Introduction
25 | P a g e
intuitive appeal to the investor community. Thus, in practice the
situation demands that substantial amount of energy needs to
be invested in quest for reasonable numbers to make logical
appeal and intuitive acceptance by manipulating the original
model. The basic motivation for framing the research is driven
by the above discussed premise.
Choosing a single asset for investment is not a difficult
task for an investor. However, the investor’s decision becomes
difficult when faced with virtually innumerable investable choices
and more importantly an infinite number of combinations of
assets. The investor chooses those portfolios which lie on the
efficient frontier. The choice may be affected by investor’s
beliefs, objectives, preferences, expectations, risk aversion, time
and budget constrains, estimations among other. In addition,
external factors will also affect the choice. Due to all the external
and internal factors, the investor faces a dynamic decision
problem in selection of optimal portfolio. Also, it is quite evident
that the interactions of risk and reward are stated in portfolio
theory in a very broad framework and has deeply influenced the
way institutional portfolios are managed, and is also successful
in motivating the “passive management” investment techniques.
The mathematics of portfolio theory is widely used in risk
Chapter 1: Introduction
26 | P a g e
management and is basic for more recent risk measures. But
when the investors use the models to solve real world problem,
every assumption of the model becomes its limitation and often
become obvious and thus is expected to have deep implications
on the actual risk and reward that the portfolio’s holders will
bear. Most recent works have shown that practically one can not
select an optimal portfolio by considering the mean-variance
portfolio theory. And above it, the gains from portfolio
optimization are seen to have been nullified by the error explicit
on the most common model’s parameters estimators (Uppal,
DeMiguel, Garlappi and Nogales, 2007). Uppal, DeMiguel,
Garlappi and Nogales (2007) have put forward their argument
by demonstrating how a naively diversified portfolio with equal
weights in every asset, can out-perform out-of-sample on a risk
adjusted basis (Sharpe-ratio in this case) an “optimally”
diversified portfolio. However, some parameters of mean-
variance model have been improved in several other studies.
Sharpe’s (1963) study is one of the examples where he
observed the market portfolio in order to improve the estimations
of the expected return and covariance matrix.
Thus, though Markowitz model is the pioneer work in
portfolio management, it suffers from some other serious
Chapter 1: Introduction
27 | P a g e
limitations. To overcome the computational complexity, it has to
rely on a number of strict technical assumptions which are more
or less away from reality viz. markets are assumed to be perfect
that means there are neither taxes nor transactions costs and
assets are infinitely divisible; investors make their decisions at
exactly one point in time for a single-period horizon; and the
means, standard deviations and correlation coefficients are
sufficient to describe the assets’ returns. This theory needs a
large number of data. W. Sharpe introduced Single Index Model
to reduce huge data need. This model can be used with
relatively few data as compared to Markowitz’s mean – variance
– efficient portfolio theory. Single Index Model assumes that the
only reason of security movement is a common co-movement
with the index and the index is unrelated to a security’s unique
return. The model does not consider other factors which affect
the security return such as company performance, economy of
the concerned industry, economic condition of the company etc.
Multi Index Model was introduced to capture some of the
non-market influences that cause securities to move together.
Non-market influences means a set of economic factors or
structural groups that account for common movement in security
prices beyond that accounted for by the market index itself. It
Chapter 1: Introduction
28 | P a g e
uses extra indices in the hope of capturing additional information
which are not present in Sharpe’s Single Index Model. The
problem of introducing additional indices is that they may pick up
random noise rather than real influences. To eliminate the
problem of picking up random noise averaging techniques were
introduced. But the disadvantage of averaging technique is that
real information may be lost in the averaging process.
In spite of the classical mean – variance- optimum
portfolio theories, some other approaches are also used
frequently by the authors such as stochastic dominance,
geometric mean return and analysis in terms of characteristics
of the return distribution etc. Later many works have been
reported in the literature on portfolio management. Many authors
have suggested selection of portfolio in many ways. These have
established a close relationship with statistic of modeling.
Various stochastic formulations (see Marton,1980; Sahalia &
Brandt, 2001; Detemple, Garcia and Rindisscher, 2003;
Beliakov and Bagirov, 2006; Okhrin and Schmid, 2008 etc.)
have been used to discus the problem of selection of portfolio
and various complicated statistical tools have been used to
discus the problem of selection of weight of the portfolio (the
proportion of wealth invested in each individual asset). However,
Chapter 1: Introduction
29 | P a g e
the adoption of more sophisticated risk measures like value at
risk and constraints including restrictions on the maximum
number of different assets in a portfolio and minimum holding
size, have made it all but impossible to optimize portfolios with
classical techniques.
Earlier attempt by Elton and Gruber (1973) and recent
attempt by Ledoit and Wolf (2004) has come up with the finding
that the output of optimally portfolios can be improved by
imposing a structure to the covariance matrix as opposed to its
sample estimator. Although this recent research gives some
insight for the mean-variance approach, it fails to give a well-
known robust estimate for the expected return of most assets
and also fails to achieve the benefits promised by portfolio
optimization in its conception. Hence, there is a need to delve
deep into the nuances of mean-variance approach to design an
integrated approach to address the issue of portfolio
optimization in more simplistic terms and churn out results from
the proposed model which will be more attractive intuitively. To
address the issue, the need to frame integrated robust model
motivates the present research to integrate different
mathematical and statistical models along with heuristics which
allow us to estimate and emulate the risk features of a given
Chapter 1: Introduction
30 | P a g e
portfolio and to use simulation techniques to generate scenarios
and weights which enable us to perform portfolio optimization
taking into account explicitly the role of value system in the
decision making framework of the investor.
1.6 Preview of the work
Selection of the optimum portfolio is a complex task for
the general investors as choice of optimum weight is very
difficult to make. There may be basically two ways of arriving at
an optimum portfolio – one by minimizing the risk and the other
by maximizing the return. This doctoral work proposes to strike a
balance between these two. In chapter 3 optimum portfolios
have been constructed subject to minimum return constraints.
This minimum return constraint starts from the minimum return
of the security in the portfolio and increases step by step to
maximum return of the security in the portfolio. Then a heuristic
procedure for arriving at security weights has been introduced
based on the investors’ propensity to take risk. For this purpose,
two extreme situations have been chosen – risk taker and risk
aversive investor. After constructing heuristic portfolio the extent
of closeness between the ideal portfolio constructed on the
basis of optimization method and portfolio constructed on the
basis of heuristic methods has been examined. For this purpose
Chapter 1: Introduction
31 | P a g e
Euclidian distance is considered. After detailed analysis, a point
of change have been identified beyond which the optimum
portfolio is closer to optimistic portfolio than to pessimistic
portfolio before which the optimum portfolio is closer to
pessimistic portfolio than optimistic portfolio.
In chapter 4 the optimum portfolio is obtained through a
mathematical programming framework so as to minimize the
portfolio risk subject to return constraint expressed in terms of
coefficient of optimism (α), where α varies from 0 to 1.
Simultaneously, four heuristic portfolios have been developed
for optimistic and pessimistic investors, risk planners and
random selectors. Given the optimum portfolio and a heuristic
portfolio City Block Distance has been calculated to measure the
departure of the heuristic solution from the optimum solution.
In chapter 5 coefficient of optimism has been introduced
in the weight of risk planner to observe the change of the
behavior of the heuristic portfolio. The City Block Distance is
used to calculate the distance between the optimum portfolio
and the heuristic portfolios. For moderate values of the
coefficient of optimism a heuristic investor’s decision nearly
coincides with the corresponding optimum portfolio. However,
Chapter 1: Introduction
32 | P a g e
for extreme situations i.e. optimistic and pessimistic situations
heuristic portfolio differs from optimum portfolio.
Chapter 6 states the comparison between Sharpe’s cut
off principle portfolio and proposed near optimum portfolio with
that of optimum portfolio under Sharpe’s Single Index Model.
Here also the coefficient of optimism in the decision making
process has been considered to compare Sharpe’s approach
under optimality principle and cut off principle and the proposed
near optimum portfolio, based on Single Index Model and to
examine the suitability of near optimum portfolio over Sharpe’s
cut off principle portfolio. These optimum portfolios have been
obtained through a mathematical programming framework so as
to minimize the portfolio risk subject to return constraint
expressed in terms of coefficient of optimism. To know the
similarity between the cut off principle portfolio and the near
optimum portfolio with that of optimum portfolio under Sharpe’s
Single Index model, City Block Distance has been considered.
Up to moderate value and very high value of coefficient of
optimism, near optimum portfolio shows better result. However,
for moderate to high value of coefficient of optimism, the cut off
principle portfolio shows closer result. This put forward the
admissibility of the near optimum portfolio.
Chapter 1: Introduction
33 | P a g e
Chapter 7 concludes the work and presents some
limitation of the present work. It also provides a direction for
future research which can be carried out to make the proposed
approach more robust and practically applicable.
1.7 Scope of the work
The portfolio optimization problem is mainly concerned
with selecting the optimal investment strategy of an investor. In
other words, the investor looks for an optimal decision on how
many shares of which security should be purchased to maximize
the expected utility. If the investor knows the securities that may
give maximum expected return or minimum expected risk, it is
easy to take optimal decision. But in real world it is difficult to
find out those securities due to presence of efficient market. The
statistical models used in behavioural finance are not very easily
understandable to the general investor. When a general investor
wishes to invest money in any portfolio of securities they are
more concerned about the expected return and risk of the
portfolio not about the various statistical models. The present
study mainly focuses on the weight of the securities in the
portfolio and proposes a simple heuristic tool to help those
investors so that they can get a near optimum portfolio for
Chapter 1: Introduction
34 | P a g e
investment. Heuristic method is not universally accepted but is
having intuitive appeal.
The present study aims to identify the objectives,
background, methodology, and proposes a model which
proposes to simplify the portfolio optimization problem. The
academic endeavor provides a rigorous treatment to the weight
as a decision variable in the optimization framework. The
decision variable obtained heuristically is also factored into the
optimization framework so as to provide an all inclusive
dimension to the simplified approach. The work also provides a
framework and analysis of the allocation decisions of the linear
programming model and non-linear programming model. The
interactions of the different value system with different
decision- making systems of the investors have been well
captured in the heuristic model generation process and its
applicability has been ensured by comparing with the
performance of the optimal solution generated by classical
models.
One can get number of values for weight, risk and return
in the optimization framework depending on the computational
techniques used to compute weight, risk and return. The
resultant weight values churned out form optimized solutions
Chapter 1: Introduction
35 | P a g e
provides a complex set which thereby makes it practically
impossible to determine the global optima. Hence, in order to
simplify the process, weights have been considered as the only
decision variable in the optimization framework considering the
assumption that for different set of investors the risk and return
are held constant but the proportion of total investment in
different securities in the portfolio can be manipulated and
hence controlled to reach at the optimum solution. Therefore,
according to the class of investors and their corresponding risk
appetite heuristic weights can be generated. Heuristic portfolios
are designed to compute the heuristic risk and return to
compare with the optimum portfolios’ risk and return to ascertain
their closeness and make informed decisions.
V{tÑàxÜ E
exáxtÜv{ `xà{ÉwÉÄÉzç
Chapter 2: Research Methodology
36 | P a g e
2.1 Introduction
Empirical research can be conducted by basically
following two approaches: quantitative and qualitative approach.
The generation of data in quantitative form and its subsequent
rigorous treatment in a formal and rigid fashion defines
quantitative approach. Quantitative approach can be further sub-
divided into inferential, experimental and simulation approaches
to research. In our study we have adopted experimental
approach whereby we have exercised much control over the
research environment and some variables are manipulated to
observe their effect on other variables. We have also adopted
simulation approach wherein construction of an artificial
environment is facilitated so that relevant information and data
can be generated and its effect can be studied. Thus, our
research approach has allowed for an observation of the
dynamic behaviour of a system (or its sub-system) under
controlled conditions.
2.2 Research Questions
The computational complexity of the optimization models
coupled with different investor sentiments and the role played by
different value systems ultimately directs to address the issue
Chapter 2: Research Methodology
37 | P a g e
and demands for designing and selecting optimum portfolio
based on different value systems. The classification of the
investors sentiments and the role played by risk taking aptitude
in portfolio optimization calls for considering the issue of why not
to consider the concept of degree of optimism in the decision
framework of the portfolio optimization. Also it is necessary to
have a mathematical formulation to arrive at the optimum
decision rule given a value for degree of optimism of the
investor. Thus we note that there is a need to suggest a
heuristic framework and subsequently examine the closeness
between optimum solution and heuristic solution. Of the
classical models of portfolio optimization Sharpe’s model is most
accepted. The basic task thus is to reconsider Sharpe’s model,
and subsequently come up with modified decision rule. Lastly it
also becomes pertinent to examine the closeness between
Sharpe’s model and modified decision rule.
2.3 Objectives
The main objective of this study is to select optimum
portfolio based on different value systems. Specifically, the
study has the following objectives:
Chapter 2: Research Methodology
38 | P a g e
1. To introduce the concept of co-efficient of optimism
in the decision framework of the portfolio
optimization.
2. To present the mathematical formulation to arrive at
the optimum decision rule given a value for
coefficient of optimism.
3. To suggest a heuristic framework and examine the
closeness between the optimum solution and
heuristic solution.
4. To reexamine Sharpe’s model.
5. To suggest a modified decision rule, based on
Sharpe’s concept.
6. To make a comparative study between Sharpe’s
model and modified decision rule.
2.4 Research Design
2.4.1 Sampling Design
For empirical study, we have considered daily security
wise data of fifteen companies listed in Nifty for ten years (2000
to 2009). Data have been collected from www.nseindia.com.
Then, random selection technique has been considered to
Chapter 2: Research Methodology
39 | P a g e
select fifteen companies out of fifty companies listed in the Nifty.
To select the companies 11th June, 2010 trading day has been
randomly considered. Jindal Steel, Ranbaxy, Reliance, HDFC,
Tata Motors, Hindalco, Wipro, Ambuja Cement, Siemens, Tata
Steel, ACC, Reliance Infrastructure, Axis Bank, BPCL and Sun
Pharmaceutical – these fifteen companies have been
considered for empirical analysis based on simple random
sampling.
As per the specifications of the model requirement and
for the purpose of model validation as discussed in chapter six
different data set has been used. The example given by Elton,
Gruber and Padberg (1978) has been considered for arriving at
the optimal portfolio as per Sharpe’s Single Index Model.
2.4.2 Statistical Design
To undertake comparative study we need to find out the
deviation from the optimum decision rule. This calls for suitable
choice of a measure of distance. In statistics, various types of
distance measures are found. The main distance measures are
discussed one by one.
Chapter 2: Research Methodology
40 | P a g e
City block distance
The City block distance between two points,
a( ~a a1,a2,a3,…….ak.) and b (
~b b1,b2,b3,……bk.), with k
dimensions is calculated as:
k
jjj ba
1 , j=1,2,3,……,k
The City block distance is always greater than or equal to
zero. The measurement would be zero for identical points and
high for points that show little similarity. It is being used in plant
location and other areas (Wikipedia, 2012).
Euclidean distance
Euclidean distance or Euclidean metric is the "ordinary"
distance between two points that one would measure with a
ruler, and is given by the Pythagorean formula. By using this
formula as distance, Euclidean space becomes a metric space.
The associated norm is called the Euclidean norm. Older
literature refers to the metric as Pythagorean metric.
Chapter 2: Research Methodology
41 | P a g e
Definition
The Euclidean distance between points p and q is the
length of the line segment connecting them ( ).
In Cartesian coordinates, if p = (p1, p2,..., pn) and
q = (q1, q2,..., qn) are two points in Euclidean n-space, then the
distance from p to q, or from q to p is given by:
(1)
Euclidean vector is the position of a point in a Euclidean
n-space. So, p and q are Euclidean vectors, starting from the
origin of the space, and their tips indicate two points. The
Euclidean norm, or Euclidean length, or magnitude of a vector
measures the length of the vector:
where the last equation involves the dot product.
A vector is a directed line segment from the origin of the
Euclidean space (vector tail), to a point in that space (vector tip).
If it is considered that the length of the vector is actually the
distance from its tail to its tip, it becomes clear that the
Chapter 2: Research Methodology
42 | P a g e
Euclidean norm of a vector is just a special case of Euclidean
distance: the Euclidean distance between its tail and its tip.
The distance between points p and q may have a
direction (e.g. from p to q), so it may be represented by another
vector, given by
The Euclidean distance between p and q is just the
Euclidean length of this distance or displacement vector (when p
and q represent two positions of the same point at two
successive instants of time):
2)
Which is equivalent to equation 1, and also to:
One dimension
In one dimension, the distance between two points on the
real line is the absolute value of their numerical difference. Thus
Chapter 2: Research Methodology
43 | P a g e
if x and y are two points on the real line, then the distance
between them is given by:
In one dimension, there is a single homogeneous,
translation-invariant metric up to a scale factor of length, which
is the Euclidean distance. In higher dimensions there are other
possible norms.
Two dimensions
In the Euclidean plane, if p = (p1, p2) and q = (q1, q2) then
the distance is given by
This is equivalent to the Pythagorean theorem.
N dimensions
In this way the distance for an n-dimensional space is
(Wikipedia, 2012)
Chapter 2: Research Methodology
44 | P a g e
Mahalanobis distance
Mahalanobis distance is a distance measure introduced
by P. C. Mahalanobis in 1936. According to this distance
different patterns can be identified and analyzed with the help of
correlation between the variables. It describes the similarity of
an unknown sample set to a known one. It takes into account
the correlations of the data set and is scale-invariant which are
not considered in case of Euclidean distance. In other words, it
is a multivariate measure
Definition
Formally, the Mahalanobis distance of a multivariate
vector from a group of values with
mean is defined as:
where, = variance-covariance matrix
Mahalanobis distance can also be defined as a
dissimilarity measure between two random vectors and of
the same distribution with a common covariance matrix as
Chapter 2: Research Methodology
45 | P a g e
The Mahalanobis distance reduces to the Euclidean
distance when the covariance matrix is the identity matrix. If the
covariance matrix is diagonal, then the resulting distance
measure is called the normalized Euclidean distance:
Where, is the standard deviation of the and over
the sample set.
The Mahalanobis distance can be used to classify a test
point as belonging to one of k classes. One first estimates the
common covariance matrix of all the classes, usually based on
samples known to belong to each class. Then the Mahalanobis
distance to each class can be computed with a given test
sample, and can be classified as belonging to that class for
which the Mahalanobis distance is minimal.
Applications
Mahalanobis' discovery was prompted by the problem of
identifying the similarities of skulls based on measurements in
1927. It is widely used in cluster analysis and classification
Chapter 2: Research Methodology
46 | P a g e
techniques. It is closely related to Hotelling's T-square
distribution used for multivariate statistical testing and Fisher's
Linear Discriminant Analysis that is used for supervised
classification. This distance can also be used in the
development of the linear regression model (Wikipedia, 2012).
Chebyshev distance
Chebyshev distance or Tchebychev distance is
developed by Pafnuty Chebyshev. It is a metric defined on a
vector space where the distance between two vectors is the
greatest of their differences along any coordinate dimension.
The distance is used in chess. The Chebyshev distance
between two spaces on a chess board gives the minimum
number of moves a king requires to move between them. For
this reason it is also known as chessboard distance.
Definition
The Chebyshev distance between two vectors or points p
and q, with standard coordinates and , respectively, is
Chapter 2: Research Methodology
47 | P a g e
hence it is also known as the L∞ metric (Wikipedia, 2012).
Minkowski distance
The Minkowski distance is a metric on Euclidean space
which can be considered as a generalization of both the
Euclidean distance and the Manhattan distance.
Definition
The Minkowski distance of order p between two points
is defined as:
The Minkowski distance is a metric as a result of the
Minkowski inequality (Wikipedia, 2012).
Reasons for considering Euclidean Distance and City Block
Distance in the Study
However, in this doctoral thesis only Euclidean Distance
(ED) and City Block Distance (CBD) have been considered as
two standard measures of distance for a formal look at the
Chapter 2: Research Methodology
48 | P a g e
behavior of heuristic solutions in respect of the optimum
solutions. In other words, with these distance measures the
similarity between the optimum choice of portfolios and the
heuristic choice of portfolios is computed. As distance between
two points has been measured in this thesis, ED and CBD are
the appropriate measures. The Mahalonobis distance is used in
cluster analysis and classification techniques mainly. Chebyshev
distance is mainly used in chess. For these reasons these
distance measures have not been used.
Decision theory
Decision theory is concerned with identifying the values,
uncertainties and other issues relevant in a given decision, its
rationality, and the resulting optimal decision. It is closely related
to the field of game theory as to interactions of agents with at
least partially conflicting interests whose decisions affect each
other.
Most of decision theory is normative or prescriptive.
Normative decision theory is concerned with identifying the
optimum decision to take, assuming an ideal decision maker
who is fully informed, able to compute with perfect accuracy,
and fully rational. When people violates the optimality rules i.e.;
Chapter 2: Research Methodology
49 | P a g e
people do not behave in ways consistent with axiomatic rules,
often their own, there is a related area of study, called a positive
or descriptive discipline. This type of decision theory is
attempting to describe what people will actually do. Since the
normative, optimal decision often creates hypotheses for testing
against actual behaviour, the two fields are closely linked.
In recent decades, a new term ‘behavioural decision
theory’ is emerging and this has contributed to a re-evaluation of
what rational decision-making requires.
Some decisions are difficult because of the need to take
into account how other people in a given situation will respond to
the decision that is taken. Game theory is used to analyse such
social decisions. From the standpoint of game theory most of
the problems treated in decision theory are one-player games
(or the one player is viewed as playing against an impersonal
background situation).
Other areas of decision theory are concerned with
decisions that are difficult simply because of their complexity, or
the complexity of the organization that has to make them. In
such cases the difficulty arises in determining the optimal
behaviour. The Club of Rome, for example, developed a model
Chapter 2: Research Methodology
50 | P a g e
of economic growth and resource usage that helps politicians
make real-life decisions in complex situations.
Alternatives to decision theory
The probability theory can be used as an alternative to
decision theory. Though, it is a highly controversial issue. The
proponents of fuzzy logic, possibility theory, Dempster–Shafer
theory and info-gap decision theory state that probability is only
one of many alternatives and point to many examples where
non-standard alternatives have been implemented with apparent
success. Non- probabilistic rules such as minimax and maximax
principles are also used in making decision. The rules are
robust. Hurwitz developed the maximin and maximax rules to
take decision (Wikipedia, 2012).
Hurwitz Criterion
A compromise between the maximax and maximin criteria
is the Hurwitz criterion. Maximax criterion assumes that the
decision maker is not optimistic. According to this criterion the
maximum payoff for each alternative is located first and then the
alternative with the maximum number will be selected. This
criterion decision locates the alternative with the highest
possible gain. The maximin criterion, on the other hand,
Chapter 2: Research Methodology
51 | P a g e
assumes that the decision maker is not totally pessimistic and in
maximin criterion, the minimum payoff for each alternative is
located first and then the alternative with the maximum number
will be selected .
Thus, the Hurwitz principle assumes that the decision
maker is neither totally optimistic nor totally pessimistic. With the
Hurwicz Criterion, the decision payoffs are weighted by a
Coefficient of Optimism, a measure of the decision maker’s
degree of optimism. The coefficient of optimism, defined as “α”,
lies between 0 and 1 (i.e. 0≤α≤1.0).
α=1.0 represents that the decision maker is completely
optimistic; α=0 represents that the decision maker is completely
pessimistic. (Given this definition, 1-α is the Coefficient of
Pessimism). When the value of α=0.5, the decision maker is
neutral. The advantage of this approach is that it allows the
decision maker to build in personal feelings about relative
optimism and pessimism.
The formula is
Criterion of Realism = α (maximum payoff for an
alternative) + (1-α) (minimum payoff for an alternative)
Chapter 2: Research Methodology
52 | P a g e
As Hurwitz principle talks about optimistic, pessimistic
and neutral investors (to be referred as risk planner in the
present study) this decision criterion is applicable in the present
study (Wikipedia, 2012).
Linear programming
Linear programming or linear optimization is a
mathematical method for determining a way to achieve the best
outcome (such as maximum profit or minimum cost) in a given
mathematical model for some list of requirements represented
as linear relationships. Linear programming is a specific case of
mathematical programming or mathematical optimization.
In other words linear programming is a technique for the
optimization of a linear objective function, subject to certain
linear constraints. The constraints may be linear equality or
linear inequality type. Its feasible region is a convex polyhedron.
That means the feasible region is a set defined as the
intersection of finitely many half spaces, each of which is
defined by a linear inequality. Its objective function is a real-
valued affine function defined on this feasible region. A linear
programming algorithm finds a point in the feasible region where
Chapter 2: Research Methodology
53 | P a g e
this function has the smallest (or largest) value if such point
exists.
Linear programs are problems that can be expressed in
canonical form:
where x represents the vector of variables (to be
determined), c and b are vectors of known coefficients, A is a
known matrix of coefficients, and is the matrix transpose.
The expression to be maximized or minimized is called the
objective function (cTx in this case). The inequalities Ax ≤ b and
X >= are the constraints which specify a convex polytope over
which the objective function is to be optimized. In this context it
can be stated that when two vectors have the same dimensions,
they can be comparable. If every entry in the first is less-than or
equal-to the corresponding entry in the second then the first
vector is less-than or equal-to the second vector.
Linear programming can be applied to various fields of
study. It is used in business, economics, finance and in some
engineering problems. Industries that use linear programming
Chapter 2: Research Methodology
54 | P a g e
models include transportation, energy, telecommunications, and
manufacturing. It is also useful in modeling diverse types of
problems in planning, routing, scheduling, assignment, and
design.
Existence of optimal solutions
Geometrically, the linear constraints define the feasible
region, which is a convex polyhedron. When a linear function is
a convex function, it implies that every local minimum is a global
minimum; similarly, when a linear function is a concave function,
it represents that every local maximum is a global maximum.
Optimal solution need not exist, for two reasons. First, no
feasible solution exists if two constraints are inconsistent.
Second, when the feasible region is unbounded in the direction
of the gradient of the objective function, which is the vector of
the coefficients of the objective function, then no optimal value is
attained.
Otherwise, in the presence of a feasible solution and
bounded objective function, the optimum value is always
attained on the boundary of optimal level-set, by the maximum
principle for convex functions (alternatively, by the minimum
principle for concave functions).However, some problems have
Chapter 2: Research Methodology
55 | P a g e
distinct optimal solutions. For example, when the objective
function is the zero function, the problem of finding a feasible
solution to a system of linear inequalities is a linear
programming problem. For this feasibility problem with the zero-
function for its objective-function, if there are two distinct
solutions, then every convex combination of the solutions is a
solution.
Algorithms
A series of linear constraints on two variables produces a
region of possible values for those variables. Solvable problems
will have a feasible region in the shape of a simple polygon. The
vertices of the polytope are also called basic feasible solutions
(Wikipedia, 2012).
Nonlinear programming
Nonlinear programming (NLP) is defined as the process
of solving a system of equalities and inequalities (collectively
termed constraints) over a set of unknown real variables, along
with an objective function to be maximized or minimized, where
at least one of the constraints and the objective function is
nonlinear.
Chapter 2: Research Methodology
56 | P a g e
Applicability
A typical non-convex problem is that of optimising
transportation costs with various connectivities and capacity
constraints. The transportation costs can be selected from a set
of transportation methods, one or more of which exhibit
economies of scale. An example would be petroleum product
transport given a selection or combination of pipeline, rail tanker,
road tanker, river barge, or coastal tank ship. Owing to
economic batch size the cost functions may have discontinuities
in addition to smooth changes.
Mathematical formulation of the problem
The problem can be stated simply as:
to maximize some variable such as product
throughput
or
to minimize a cost function
where
Chapter 2: Research Methodology
57 | P a g e
s.t. (subject to)
Methods for solving the problem
If the objective function f is linear and the constrained
space is a polytope, the problem is a linear programming
problem and can be solved by using linear programming
solutions.
If the objective function is concave or maximization type
problem, or convex or minimization type problem and the
constraint set is convex, then the program is called convex and
then convex optimization method can be used to solve the
problem.
When the objective function is a ratio of a concave and a
convex function (in the maximization case) and the constraints
are convex, then fractional programming techniques can be
used by transforming the problem to a convex optimization
problem.
Chapter 2: Research Methodology
58 | P a g e
To solve non-convex problems several methods are
available. One approach is to use special formulations of linear
programming problems. Another method involves the use of
branch and bound techniques. In this method the program is
divided into subclasses to be solved with convex (minimization
problem) or linear approximations that form a lower bound on
the overall cost within the subdivision. With subsequent
divisions, at some point an actual solution will be obtained
whose cost is equal to the best lower bound obtained for any of
the approximate solutions. This solution is optimal, although
possibly not unique. The best possible solution is within a
tolerance from the best point found, known as ε-optimal.
Terminating to ε-optimal points is typically necessary to ensure
finite termination. This ε-optimal solution is useful for large,
difficult problems and problems with uncertain costs or values
where the uncertainty can be estimated with an appropriate
reliability estimation. Under differentiability and constraint
qualifications, the Karush–Kuhn–Tucker (KKT) conditions give
necessary conditions for a solution to be optimal. These
conditions are also sufficient under convexity (Wikipedia, 2012).
Chapter 2: Research Methodology
59 | P a g e
2.5 Methods to be used
2.5.1 Optimization Method
Every investor fundamentally wants to select an optimal
portfolio. But construction of optimal portfolio is very difficult. The
basic of portfolio optimization method has given by Markowitz in
1952. Optimization of portfolio can be done in two ways – (1)
maximizes return subject to a certain level of risk and (2)
minimize risk subject to a certain level of minimum return. An
investor always wants to go for either of these above two
optimizations.
The present thesis has considered Markowitz
optimization model. The objective function is considered to
minimize the risk subject to a step by step increase in the return.
The two optimization approaches can be balanced by
minimizing the risk subject to a step-by-step increase in the level
of minimum return. For the lower values of minimum return the
risk can be minimized into lesser restrictions. For higher values
of the minimum return the minimization problem becomes more
restrictive giving thereby more importance on return.
In the true sense, choice of portfolio depends on the
value system of the investors and arriving at value based
Chapter 2: Research Methodology
60 | P a g e
optimum solution may not be an easy task. To handle this
problem, the thesis proposes to incorporate in the optimization
framework the coefficient of optimism (α) introduced by Hurwitcz
(see Taha) in the context of decision theory.
Depending on the choice of an optimum solution can be
obtained in a non linear programming approach. One can get
the optimum portfolio corresponding to the minimum return of
the security by considering = 0 and the optimum portfolio
corresponding to the maximum return of the security by
considering =1. Here, = 0 represents pessimistic investors
and = 1 represents optimistic investors. For random selector
and risk planner’s the value of lies between 0 and 1. In each
case, one can get different optimum weights and with those
weights return and risk of the optimum portfolio can be
calculated.
2.5.2 Heuristic Method
Classical optimization models may not be easy to handle
in the real world. Sometimes, heuristic methods are the only
way out. Mansini and Speranza (1999), Chang et al. (2000)
suggested the use of heuristic optimization techniques to
portfolio selection. Heuristic defines as experience-based
Chapter 2: Research Methodology
61 | P a g e
techniques for problem solving, learning, and discovery.
Heuristic methods are used to speed up the process of finding
an acceptable solution where an in-depth search is not possible.
In other words, heuristic methods are the strategies which are
readily accessible but loosely applicable. A rule of thumb, an
educated guess, an intuitive judgment, or common sense etc. is
used as the heuristic methods.
In the present thesis the security weights of the investors
can also be generated heuristically based on the value system
of the investors. Four types of investors are considered
according to their propensity to take risk – risk taker or optimistic
investor, risk averse or pessimistic investor, risk planner and
random selector. As their risk apatite is different, their security
weights are different.
Optimistic investor’s put emphasis on the expected
return. So when the weight of optimistic investor is to be
generated emphasis has been given to the security returns and
the security weights are considered as directly proportional to
the expected return of the security. For heuristic pessimistic
portfolio, security weights are to be considered as inversely
proportional to standard deviation of the security. Here an
assumption is to be made that a pessimistic investor gives more
Chapter 2: Research Methodology
62 | P a g e
importance on risk of the portfolio ignoring the expected return
part and more is the risk less is the chance of selection. On the
other hand, as the risk planner considers both risk and return in
choosing the securities of a portfolio, the formula of security
weight for them is to be constructed by considering both return
and risk of the securities. Lastly for random selector the
assumption is to be taken that they randomly select securities
for the portfolio instead of giving importance to either return or
risk. Thus, they put equal weight to each security in the portfolio.
2.5.3 Model based
Prior to Markowitz's work, the assessment of the risks
and rewards of portfolios was carried out through the analysis of
individual securities independently. Markowitz was the person
who first introduced the concept of diversification of securities.
He proposed that investors should focus on joint risk-reward
features of the securities while selecting portfolios. He argued
that a portfolio consisted with more than one security would
have less risk than that of an individual security.
Markowitz assumed that, a rational investor would
choose the portfolio with the highest expected return, for a given
Chapter 2: Research Methodology
63 | P a g e
level of risk and similarly, for a given level of expected return, a
rational investor would choose the portfolio with the lowest risk.
The theory used historical returns of each asset of the
portfolio and average return is used to estimate expected return.
Standard deviation or variance of return is used to calculate risk
and linear correlation is also used for this purpose. Markowitz
used volatility and expected return as proxies for risk and
reward.
Markowitz defined an “optimal” way of selecting a
portfolio by balancing the risk and reward features of the
portfolio within the infinite possible alternatives of portfolio. In
other words, a portfolio is said to be optimal if there is no
portfolio having the same risk with a greater expected return,
and there is no portfolio having the same expected return with a
lower risk. The set of portfolios constructed in this optimal
manner conform what he called the efficient frontier. He
concludes that an investor should select a portfolio that lies on
the efficient frontier.
The classical Markowitz portfolio selection is based on
the assumption of multinormally distribution of returns. Since the
multinormal model is inadequate for risk estimation, the
Chapter 2: Research Methodology
64 | P a g e
Markowitz optimal portfolio might not be optimal any more. The
mean-variance theory is applicable where the asset class is
different and it is giving an open form of solution.
The Single Index Model is based on the observation that
securities move together with the market. That means when a
market represented by a market index moves up or down,
security prices will also move accordingly. Therefore, there exist
a positive correlation between the security prices and the market
index. In fact, the Single Index Model assumes that the only
relevant cause of correlation between two assets is their
common dependency on market movements. The other
assumption of the model is that a security’s return does not
affect the market index. That means, the index is unrelated to
unique return.
The mathematical model of Single Index Model is a two
variable linear regression equation with the return on each
security as dependent variable and the return on a market index
as independent variable. Like Markowitz, Sharpe also used
variance as the measure of risk. But he introduced covariance
term to calculate risk of a portfolio.
V{tÑàxÜ F
fxÄxvà|ÉÇ Éy cÉÜàyÉÄ|É Utáxw ÉÇ `|Ç|ÅâÅ exàâÜÇ VÉÇáàÜt|Çà
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
65 | P a g e
3.1 Introduction
A portfolio means a combination of two or more assets.
From a given set of assets a large number of portfolios can be
selected. Each portfolio has its own risk-return characteristics. In
view of the innumerable parameters involved in selection of a
portfolio the concerned decision making becomes a complicated
task. Selection of the optimum portfolio is a complex task for the
general investors as choice of optimum weight is very difficult to
make. In most of the cases it is difficult to arrive at an optimal
solution. One of the key problems of portfolio optimization is to
obtain correct and precise portfolio weights (the function of
wealth invested in individual asset). Precise estimation of
weights reduces unnecessary transaction costs and helps to
construct the optimum portfolio. The problem of optimum
portfolio selection has been discussed by various researchers
from various angles. Some authors preferred classical Mean –
Variance technique to select optimal portfolio, whereas other
group proposed new methods of optimal portfolio selection.
Arriving at a heuristic solution may be possible at a given point
of time and at a given set of conditions but the same may not be
a uniformly optimal one.
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
66 | P a g e
When Markowitz (1952) developed a Mean – Variance
framework of modern portfolio optimization more that 50 years
ago, simplicity was the key part of its appeal. However, the
adoption of more sophisticated risk measures like value at risk
and constraints including restrictions on the maximum number of
different assets in a portfolio and minimum holding size, have
made it all but impossible to optimize portfolios with classical
techniques. Financial analysts have been talking about
downside risk for years but few have attempted to use it for
portfolio selection because of the resultant optimization problem
and its difficulty to solve. Since the crucial work of Markowitz
and Sharpe Mean – Variance have been the central focuses of
portfolio analysis, most authors follow this Mean – Variance
technique to select an optimum portfolio. The present study has
tried to show that one investor heuristically can reach near to the
optimum solutions.
3.2 Earlier works
Ross (1976) developed an alternative asset pricing
theorem called Arbitrage Pricing Theory (APT) which is based
on less restrictive assumptions. Under no arbitrage conditions,
the theory states that the return on any stock is linearly related
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
67 | P a g e
to asset of systematic factor or risk factors. It means that two
portfolios that have the same risk cannot offer different expected
returns. However, APT fails to prescribe the nature and number
of factors.
Most asset allocation analysis used the Mean – Variance
approaches for analyzing the trade off between risk and
expected return. Merton (1980) described the difficulty of
precisely estimating the expected portfolio return. In the
classical Merton problem the investor can allocate her/ his
money into a risk less account governed by a deterministic
interest rate and different risky stock. Merton was able to reduce
the portfolio problem to a stochastic optimum control problem,
which can be solved by using stochastic control methodology.
He considered New York stock exchange data from 1926 to
1978 for the analysis.
Jauch and Glueck (1988) have pointed out seven values.
The most prominent is risk value that affects the business
decision-making process.
Sahalia & Brandt (2001) studied the asset allocation
problem when the conditional moments of return are partly
predictable. They examined how the optimal portfolio weights
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
68 | P a g e
depend on the predictive variables. To know the nature of
dependency of the optimal portfolio weights on the predictive
variables they combined the predictor in to a single index that
best captures time variations in investment opportunities.
Schaerf (2002) discussed the problem of a portfolio
selection, which provide a suitable balance of expected risk and
return to the investors. He considered additional constraints on
the cardinality of the portfolio and on the quantity of individual
assets over the seminal mean – variance model of Markowitz.
The Detemple, Garcia and Rindisscher (2003) proposed
a new simulation based approach for the optimal portfolio
allocation in realistic environment with complex dynamics for the
state variables and large numbers of factors and assets. They
also analyzed settings where stock returns are predicted by
dividend yields and where investors have wealth dependent
relative risk aversion. They used Nasdaq index, and S& P 500
index as empirical evidence.
Beliakov and Bagirov (2006) examined numerical
performance of various methods of calculations of the
conditional values at risk (CVaR) and carried out portfolio
optimization with respect to this risk measure. They found that
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
69 | P a g e
nonsmooth optimization can be used efficiently for large portfolio
optimization and also examined parallel execution of this
method on computer cluster.
Brandt and Clara (2006) presented a novel approach to
dynamic portfolio selection that is as easy to implement as the
static Markowitz paradigm. They expanded the set of asset and
statistically optimized the portfolio in that extent asset space.
They considered conditional portfolio and timing portfolios. They
concluded that the static choice of this managed portfolios
represents a dynamic strategy that closely approximates the
optimal dynamic strategy for horizons up to five years.
A multivariate shrinkage estimator was proposed by
Okhrin and Schmid (2008) to calculate the optimal portfolio
weights. They discussed the estimated shrinkage weights
analytically and also used Monte Carlo simulation technique to
calculate the estimated shrinkage weight of the portfolio. They
have shown how the shrinkage estimator gives the most
accurate estimation of the portfolio weights. They also carried
out an empirical analysis in support of their observation.
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
70 | P a g e
3.3 Optimization Framework
In portfolio theory expected return and risk play vital role
in determining the choice set. The formula, we have used to
calculate return is given below:
ijR (Closing price of the ith security for the jth period –
Opening price of the ith security for the jth period) / (Opening
price of the ith security for the jth period) (1)
The risk of a security is the standard deviation of the
observed returns and is given by the formula:
k
jiiji R
1
2 (2)
Average return and risk for the securities so selected
have been calculated for subsequent analysis. An investor
always wants to go for either maximizing return subject to
certain level of risk or minimizing risk subject to certain level of
return. A balance between these two approaches has been
made by minimizing the risk subject to a step-by-step increase
in the level of minimum return. The optimization problem
considered for the current work is of the type:
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
71 | P a g e
Min ww ˆˆ
Subject to oiRw ˆ
μmin μ0 μmax
The risk and return of the optimum portfolio have been
calculated by using the following equations.
i
n
i
OP RwR O
i
1
(3)
OP =
n
i
n
j
ijjiji rww OO
1 1
(4)
For the numerical study the minimum return of the
security in the portfolio is considered as the starting point and in
each step the value of minimum return requirement has been
step by step increased by 0.001 point. In this process finally, the
maximum return of the security in the portfolio has been
reached. In each case, we have got different optimum weights.
With those weights return and risk of the optimum portfolio have
been calculated. Even if multiple assets are there in the
portfolio, the final return cannot be less than the minimum return
and more than the maximum return. As a result, there is no trial
and error in this approach. It is a systematic process followed
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
72 | P a g e
from minimum possibility to maximum possibility. The increment
of 0.001 is of course a subjective choice.
3.4 Heuristic Framework
In this phase a heuristic solution have been introduced
with the help of which one may get an alternative choice of
portfolio based on propensity to take risk. In this chapter, two
types of portfolios have been constructed one for optimistic
investor and the other for pessimistic investor.
In case of heuristic optimistic portfolio, we have
considered security weights directly proportional to expected
return of the security based on the assumption that an optimistic
investor gives more importance on expected return irrespective
of the risk. To ensure non-negativity in the system a locational
shift has been undertaken with point of origin as μmin. Thus, for
heuristic optimistic solution following formula has been used for
calculating the weights of the securities of the portfolio.
min1
min
i
n
i
iHOiw (5)
Given this weight function we have calculated the
expected return and risk as follows:
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
73 | P a g e
i
n
i
HOP RwR HO
i
1
(6)
HOP =
n
i
n
j
ijjiji rww HOHO
1 1
(7)
For heuristic pessimistic portfolio, security weights have
been considered as inversely proportional to standard deviation
of the security. Here we have made an assumption that a
pessimistic investor gives more importance on risk of the
portfolio ignoring the expected return part and more is the risk
less is the chance of selection. For pessimistic solution following
formula has been used for calculating the weights of the
securities of the portfolio. For heuristic pessimistic solution
following formula has been used for calculating the weights of
the securities of the portfolio.
i
n
i
iHPiw
61
61
1
(8)
Given this weight function the expected return and risk
are calculated as follows:
i
n
i
HPP RwR HP
i
1
(9)
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
74 | P a g e
HPP =
n
i
n
j
ijjiji rww HPHP
1 1
(10)
3.5 Closeness between Heuristic and Optimization
framework
The closeness between the optimum portfolio and the
heuristic portfolio is measured with the help of Euclidean
Distance, a distance between two points. The closeness is
measured to observe that whether the heuristic choice of
portfolio is near to the optimum portfolio.
3.6 Data Analysis and Interpretation
The objective is to see the closeness between the
optimum solution and the two risk-based heuristic solutions. This
has been carried out graphically to examine the closeness
between the optimum solution and heuristic solutions. For this
purpose, different tables have been framed and corresponding
graphs have been drawn for depicting the closeness between
the optimum solution and heuristic solutions. Another graph has
been drawn to find out the threshold point beyond which the
optimum portfolio is closer to optimistic portfolio than to
pessimistic portfolio and before which the optimum portfolio is
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
75 | P a g e
closer to pessimistic portfolio than optimistic portfolio. This point
has been determined by identifying the point of intersection
between lines obtained by joining the Euclidian Distance (ED)
between the optimum solution and heuristic solutions. However,
ED has been considered as a standard measure of distance for
a formal look at the behavior of heuristic solutions in respect of
the optimum solutions.
Equations 3 and 4 are used to calculate the return and
risk of the optimum portfolio respectively. Equations 5 and 8 are
used to calculate the weights of the optimistic portfolio and
pessimistic portfolio respectively. The return and risk of the
heuristic optimistic portfolio are calculated by using Equations 6
and 7 respectively. On the other hand Equations 9 and 10 are
used to determine the return and risk of heuristic pessimistic
portfolios respectively. The weights of the optimum portfolios
and heuristic portfolios under different values of minimum return
constraints, expected return and risk of the optimum portfolio,
heuristic optimistic portfolio and heuristic pessimistic portfolio
and ED between the optimum portfolios and the heuristic
portfolios have been shown separately in different tables from
Table 3.1 to Table 3.17. Regarding the graphical presentation of
closeness between optimum solutions and heuristic solutions, in
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
76 | P a g e
the horizontal axis we have measured the expected return and
in the vertical axis we have measured the expected risk. From
graph no: 3.1 to 3.8, point O represents the optimum choice of
the portfolio, point A represents the pessimistic choice of the
portfolio and point B represents the optimistic choice of the
portfolio.
The first minimum return constraint considered is -0.0057.
Table 3.1 and Table 3.2 describe the optimum weights, heuristic
optimistic weights, heuristic pessimistic weights and the returns
and risks of the three portfolios under the three different value
systems respectively. The expected return and risk of the three
portfolios are graphically described in Figure 3.1.
Table 3.1: Weights of optimum portfolio and heuristic
portfolios under -0.0057 minimum return constraint
Securities Optimum weights
Heuristic Optimistic Weights
Heuristic Pessimistic
Weights JINDALSTEL
0 0.090668 0.053313
RANBAXY
0.1936191 0.066351 0.074502
RELIANCE
0.06280846 0.077863 0.076569
HDFC
0.1424088 0.101525 0.075374
TATAMOTORS
0.005869277 0.059956 0.065455
HINDALCO
0.05411821 0.072906 0.065483
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
77 | P a g e
WIPRO
0 0.064353 0.059075
AMBUJACEM
0.1068151 0.051962 0.077734
SIEMENS
0.07751241 0.073386 0.067679
TATASTEEL
0 0.056439 0.064709
ACC
0.02198061 0.061715 0.071019
RELINFRA
0 0.064433 0.059468
AXISBANK
0.04030778 0 0.054746
BPCL
0.1195407 0.071308 0.064254
SUNPHARMA
0.1750195 0.087136 0.07062
Table 3.2: Expected return and risk of three different
portfolios under – 0.0057 as the minimum return constraint
Portfolio Expected return Expected risk
Optimum portfolio -0.001142735 0.024383995
Heuristic Optimistic portfolio
-0.001074532 0.016490497
Heuristic pessimistic portfolio
-0.001478186 0.016463641
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
78 | P a g e
Figure 3.1: Expected return and risk under -0.0057 minimum
return constraint
From Figure 3.1, it is clear that the optimum portfolio (O) and
optimistic portfolio (B) are closer to each other as compared to
the pessimistic portfolio (A). Expected returns of the optimum
portfolio and optimistic portfolio are very close. Though, the
expected risk of the optimum portfolio is higher than the risk of
optimistic portfolio. Whereas, the risks of pessimistic and the
optimum portfolio are more or less same but expected return of
optimistic portfolio is higher than that of pessimistic portfolio.
The next minimum return constraint considered is -
0.0047. Table 3.3 and Table 3.4 describe the optimum weights,
heuristic optimistic weights, heuristic pessimistic weights and the
returns and risks of the three portfolios under the three different
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
79 | P a g e
value systems respectively. The expected return and risk of the
three portfolios are graphically described in Figure 3.2.
Table 3.3: Weights of optimum portfolio and heuristic
portfolios under – 0.0047 as the minimum return constraint
Securities Optimum weights
Heuristic Optimistic Weights
Heuristic Pessimistic
Weights JINDALSTEL
0 0.090668 0.053313
RANBAXY
0.1936189 0.066351 0.074502
RELIANCE
0.06280945 0.077863 0.076569
HDFC
0.1424077 0.101525 0.075374
TATAMOTORS
0.005868386 0.059956 0.065455
HINDALCO
0.05411870 0.072906 0.065483
WIPRO
0 0.064353 0.059075
AMBUJACEM
0.1068156 0.051962 0.077734
SIEMENS
0.07751275 0.073386 0.067679
TATASTEEL
0 0.056439 0.064709
ACC
0.02197968 0.061715 0.071019
RELINFRA
0 0.064433 0.059468
AXISBANK
0.04030873 0 0.054746
BPCL
0.1195404 0.071308 0.064254
SUNPHARMA
0.1750197 0.087136 0.07062
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
80 | P a g e
Table 3.4: Expected return and risk of three different
portfolios under – 0.0047 as the minimum return constraint
Portfolios Expected Return Expected Risk
Optimum portfolio -0.00114274 0.015233167
Heuristic Optimistic
portfolio
-0.001074532 0.016490497
Heuristic
pessimistic
portfolio
-0.001478186 0.016463641
Figure 3.2: Expected return and risk under -0.0047 minimum
return constraint
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
81 | P a g e
From Figure 3.2, it is evident that the optimistic portfolio
(B) and pessimistic portfolio (A) are placed more or less same
distance to the optimum portfolio (O). The expected risk of the
optimum portfolio is much lower than that of optimistic and
pessimistic portfolios. Whereas, expected return of optimistic
portfolio is slightly higher than that of the optimum portfolio.
Optimistic portfolio’s expected return is lowest in this case. On
the other hand, expected risks of pessimistic and optimistic
portfolios are more or less same under this minimum return
constraint.
The third minimum return constraint used is -0.0037 and
Table 3.5 and Table 3.6 represent the optimum weights,
heuristic optimistic weights, heuristic pessimistic weights and the
returns and risks of the three portfolios under the three different
value systems respectively. The expected return and risk of the
three portfolios are described in Figure 3.3.
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
82 | P a g e
Table 3.5: Weights of optimum portfolio and heuristic
portfolios under – 0.0037 as the minimum return constraint
Securities Optimum weights
Heuristic Optimistic Weights
Heuristic Pessimistic
Weights JINDALSTEL
0 0.090668 0.053313
RANBAXY
0.1936181 0.066351 0.074502
RELIANCE
0.06281429 0.077863 0.076569
HDFC
0.1424053 0.101525 0.075374
TATAMOTORS
0.005865420 0.059956 0.065455
HINDALCO
0.05412018 0.072906 0.065483
WIPRO
0 0.064353 0.059075
AMBUJACEM
0.1068162 0.051962 0.077734
SIEMENS
0.07751338 0.073386 0.067679
TATASTEEL
0 0.056439 0.064709
ACC
0.02197784 0.061715 0.071019
RELINFRA
0 0.064433 0.059468
AXISBANK
0.04031004 0 0.054746
BPCL
0.1195388 0.071308 0.064254
SUNPHARMA
0.1750204 0.087136 0.07062
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
83 | P a g e
Table3.6: Expected return and risk of three different
portfolios under – 0.0037 as the minimum return constraint
Portfolios Expected Return Expected Risk
Optimum portfolio -0.00114274 0.015233165
Heuristic Optimistic
portfolio
-0.001074532 0.016490497
Heuristic
pessimistic portfolio
-0.001478186 0.016463641
Figure 3.3: Expected return and risk under -0.0037 minimum
return constraint
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
84 | P a g e
In Figure 3.3 again the optimum solution (O), the
pessimistic solution (A) and the optimistic solution (B) are
situated in the same place as under -0.0047 minimum return
constraint. Pessimistic portfolio shows the worst result in this
figure. Both expected return and risk of pessimistic portfolio are
more than that of optimistic portfolio.
The next minimum return constraint is -0.0027 and the
weights of the three different portfolios along with the expected
returns and risks of them are shown in Table 3.7 and Table 3.8
with corresponding graph (Figure 3.4) representing the position
of the three portfolios with their expected return and risk.
Table 3.7: Weights of optimum portfolio and heuristic
portfolios under – 0.0027 as the minimum return constraint
Securities Optimum weights
Heuristic Optimistic Weights
Heuristic Pessimistic
Weights JINDALSTEEL
0 0.090668 0.053313
RANBAXY
0.1936165 0.066351 0.074502
RELIANCE
0.06282562 0.077863 0.076569
HDFC
0.1423984 0.101525 0.075374
TATAMOTORS
0.005858875 0.059956 0.065455
HINDALCO
0.05412454 0.072906 0.065483
WIPRO 0 0.064353 0.059075
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
85 | P a g e
AMBUJACEM
0.1068160 0.051962 0.077734
SIEMENS
0.07751463 0.073386 0.067679
TATASTEEL
0 0.056439 0.064709
ACC
0.02197602 0.061715 0.071019
RELINFRA
0 0.064433 0.059468
AXISBANK
0.04031180 0 0.054746
BPCL
0.1195347 0.071308 0.064254
SUNPHARMA
0.1750229 0.087136 0.07062
Table 3.8: Expected return and risk of three different portfolios under – 0.0027 as the minimum return constraint
Portfolios Expected return Expected risk
Optimum portfolio -0.001142751 0.01523316
Heuristic Optimistic
portfolio
-0.001074532 0.016490497
Heuristic
pessimistic portfolio
-0.001478186 0.016463641
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
86 | P a g e
Figure 3.4: Expected return and risk under -0.0027 minimum
return constraint
Figure 3.4 projects that the expected return of the
pessimistic solution (A) is less than that of the optimum solution
(O) and risk of pessimistic solution (A) and optimistic solution
(B), both is more than that of the optimum solution (O). But the
expected return of the optimistic solution (B) is slightly more
than that of the optimum solution (O). The behaviour of the
optimum portfolio, optimistic portfolio and pessimistic portfolio
are almost same like under the minimum return constraint -
0.0037.
Now consider the minimum return constraint -0.0017. The
weights of the three different portfolios and returns and risks of
them are shown in Table 3.9 and Table 3.10 with corresponding
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
87 | P a g e
graph, Figure 3.5, representing the position of the three
portfolios under minimum return constraint -0.0017.
Table 3.9: Weights of optimum portfolio and heuristic
portfolios under -0.0017 as the minimum return constraint
Securities Optimum weights
Heuristic Optimistic Weights
Heuristic Pessimistic
Weights JINDALSTEL
0 0.090668 0.053313
RANBAXY
0.1935911 0.066351 0.074502
RELIANCE
0.06289563 0.077863 0.076569
HDFC
0.1423614 0.101525 0.075374
TATAMOTORS
0.005813745 0.059956 0.065455
HINDALCO
0.05418887 0.072906 0.065483
WIPRO
0 0.064353 0.059075
AMBUJACEM
0.1068153 0.051962 0.077734
SIEMENS
0.07754785 0.073386 0.067679
TATASTEEL
0 0.056439 0.064709
ACC
0.02203039 0.061715 0.071019
RELINFRA
0 0.064433 0.059468
AXISBANK
0.04028770 0 0.054746
BPCL
0.1195005 0.071308 0.064254
SUNPHARMA
0.1749674 0.087136 0.07062
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
88 | P a g e
Table 3.10: Expected return and risk of three different
portfolios under – 0.0017 as the minimum return constraint
Portfolios Expected return Expected risk
Optimum portfolio -0.00114272 0.015232854
Heuristic Optimistic
portfolio
-0.001074532 0.016490497
Heuristic
pessimistic portfolio
-0.001478186 0.016463641
Figure 3.5: Expected return and risk under -0.0017 minimum
return constraint
Under -0.0017 minimum return constraint the positions of
the optimum solution (O), pessimistic solution (A) and optimistic
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
89 | P a g e
solution (B) are same as under -0.0027 minimum return
constraint. The pessimistic portfolio shows the worst result.
The Table 3.11 and Table 3.12 under -0.0007 minimum
return constraint depicts the optimum weights, heuristic
optimistic weights, heuristic pessimistic weights and the returns
and risks of the three portfolios under the three different value
systems. The expected return and risk of the three portfolios are
described in Figure 3.6.
Table 3.11: weights of optimum portfolio and heuristic
portfolios under -0.0007 as the minimum return constraint
Securities Optimum weights
Heuristic Optimistic Weights
Heuristic Pessimistic
Weights JINDALSTEL
0 0.090668 0.053313
RANBAXY
0.1712044 0.066351 0.074502
RELIANCE
0.08403390 0.077863 0.076569
HDFC
0.2177282 0.101525 0.075374
TATAMOTORS
0 0.059956 0.065455
HINDALCO
0.05253162 0.072906 0.065483
WIPRO
0 0.064353 0.059075
AMBUJACEM
0.04933059 0.051962 0.077734
SIEMENS
0.08757217 0.073386 0.067679
TATASTEEL 0 0.056439 0.064709
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
90 | P a g e
ACC
0.01605094 0.061715 0.071019
RELINFRA
0 0.064433 0.059468
AXISBANK
0 0 0.054746
BPCL
0.1197338 0.071308 0.064254
SUNPHARMA
0.2018143 0.087136 0.07062
Table 3.12: Expected return and risk of three different
portfolios under – 0.0007 as the minimum return constraint
Portfolios Expected return Expected risk
Optimum portfolio -0.0007 0.015381152
Heuristic Optimistic portfolio
-0.001074532 0.016490497
Heuristic pessimistic portfolio
-0.001478186 0.016463641
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
91 | P a g e
Figure 3.6: Expected return and risk under -0.0007 minimum
return constraint
In Figure 3.6 the optimum solution (O) is going towards
the origin. That means its expected return is increasing and risk
is more or less same as under -0.0017 minimum return
constraint. From the figure it can be obtained that distance
between the optimum portfolio and the pessimistic portfolio is
more than that of the optimum portfolio and the optimistic
portfolio.
The Table 3.13 and Table 3.14 under 0.0003 minimum
return constraint depicts the optimum weights, heuristic
optimistic weights, heuristic pessimistic weights and the returns
and risks of the three portfolios under the three different value
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
92 | P a g e
systems respectively. The expected return and risk of the three
portfolios are described in Figure 3.7.
Table 3.13: Weights of optimum portfolio and heuristic
portfolios under 0.0003 as the minimum return constraint
Securities Optimum weights
Heuristic Optimistic Weights
Heuristic Pessimistic
Weights JINDALSTEL
0.1072901 0.090668 0.053313
RANBAXY
0 0.066351 0.074502
RELIANCE
0.02598263 0.077863 0.076569
HDFC
0.60152040 0.101525 0.075374
TATAMOTORS
0 0.059956 0.065455
HINDALCO
0 0.072906 0.065483
WIPRO
0 0.064353 0.059075
AMBUJACEM
0 0.051962 0.077734
SIEMENS
0 0.073386 0.067679
TATASTEEL
0 0.056439 0.064709
ACC
0 0.061715 0.071019
RELINFRA
0 0.064433 0.059468
AXISBANK
0 0 0.054746
BPCL
0 0.071308 0.064254
SUNPHARMA
0.2652069 0.087136 0.07062
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
93 | P a g e
Table 3.14: Expected return and risk of three different
portfolios under 0.0003 as the minimum return constraint
Portfolios Expected return Expected risk
Optimum portfolio 0.0003 0.020345017
Heuristic Optimistic
portfolio
-0.001074532 0.016490497
Heuristic pessimistic
portfolio
-0.001478186 0.016463641
Figure 3.7: Expected return and risk under 0.0003 minimum
return constraint
Here, the optimum portfolio has sifted to the first
quadrant. It is very clear from the above figure that the optimistic
portfolio (B) is situated near to the optimum portfolio (O) as
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
94 | P a g e
compared to the pessimistic portfolio (A). The expected return of
the optimistic portfolio is near to that of the optimum portfolio.
Expected return of the pessimistic portfolio is far from that of the
optimum portfolio. But the risks of both optimistic and
pessimistic portfolio are same.
The last minimum return constraint considered is
0.00065. The weights of fifteen selected companies for three
different portfolios are given in Table 3.15 and the expected
return and risk of the concerned portfolios have been shown in
Table 3.16. The expected return and risk of the three portfolios
are shown graphically in Figure 3.8 under 0.00065 minimum
return constraint.
Table 3.15: Weights of optimum portfolio and heuristic
portfolios under 0.00065 as the minimum return constraint
Securities Optimum weights
Heuristic Optimistic Weights
Heuristic Pessimistic
Weights JINDALSTEL
0 0.090668 0.053313
RANBAXY
0 0.066351 0.074502
RELIANCE
0 0.077863 0.076569
HDFC
1 0.101525 0.075374
TATAMOTORS
0 0.059956 0.065455
HINDALCO
0 0.072906 0.065483
WIPRO 0 0.064353 0.059075
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
95 | P a g e
AMBUJACEM
0 0.051962 0.077734
SIEMENS
0 0.073386 0.067679
TATASTEEL
0 0.056439 0.064709
ACC
0 0.061715 0.071019
RELINFRA
0 0.064433 0.059468
AXISBANK
0 0 0.054746
BPCL
0 0.071308 0.064254
SUNPHARMA
0 0.087136 0.07062
Table 3.16: Expected return and risk of three different
portfolios under 0.00065 as the minimum return constraint
Portfolios Expected return Expected risk
Optimum portfolio 0.00065 0.027169
Heuristic Optimistic
portfolio
-0.001074532 0.016490497
Heuristic pessimistic
portfolio
-0.001478186 0.016463641
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
96 | P a g e
Figure 3.8: Expected return and risk under 0.00065
minimum return constraint
Under 0.00065 minimum return constraint, again the
optimum solution is going far from the other two heuristic
solutions. Though from the above figure it is clear that the
optimistic solution (B) is situated near to the optimum solution
(O) but the pessimistic solution (A) is situated far from the
optimum solution (O). The expected return of optimistic portfolio
is more than that of pessimistic portfolio. But the expected risks
of the optimistic portfolio and pessimistic portfolio are more or
less same.
Now the ED values under all minimum return constraints
are shown in Table 3.17 to graphically present the curves drawn
by considering the coordinates between the minimum return
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
97 | P a g e
constraints and the ED between the optimum solution and the
heuristic solutions. Figure 3.9 shows the said curves. To present
Euclidian Distance (ED) graphically the horizontal axis has
measured the minimum return constraints and the vertical axis
has measured the ED values between the optimum and
pessimistic solutions and the ED values between the optimum
and optimistic solutions.
Table 3.17: Euclidian Distance between pessimistic and
optimum solutions and between optimistic and optimum
solutions
Minimum Return
Constraints
Ed between
optimum and
Pessimistic
solutions
ED between
optimum and
Optimistic solutions
-0.0057 0.0079274545032 0.0078937946135
-0.0047 0.0012753798258 0.0012591483233
-0.0037 0.0012753798258 0.0012591483233
-0.0027 0.0012753823348 0.0012573369874
-0.0017 0.001275685698 0.0012594901746
-0.0007 0.0013331751198 0.0011708631432
0.0003 0.0042693120289 0.0040922686373
0.00065 0.0109148470901 0.0108168589202
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
98 | P a g e
Figure 3.9: ED of optimistic and pessimistic solutions
In Figure 3.9, the point T1 and T2 are the two threshold points.
This identification of point of change has been undertaken for
the two risk situations – risk taker or optimistic investor and risk
aversive or pessimistic investor. Before the point T1 the ED
between the optimum solution and the pessimistic solution and
the ED between the optimum solution and optimistic solution are
same. They are overlapping before point T1. But beyond the
point T1 the ED between the optimum solution and the optimistic
solution is less than the ED between the optimum solution and
pessimistic solution. In other words, before the threshold point
T1, both the pessimistic approach and the optimistic approach
are closer to the optimization approach. The situation has got
change beyond the point T1. We have got the same result
beyond the second threshold point T2. Beyond the point T2 the
Chapter 3: Selection of Portfolio based on Minimum Return Constraint
99 | P a g e
pessimistic solution and the optimistic solution are overlapping.
In between the points T1 and T2, an optimistic investor can get
the benefit of the optimum solution.
3.7 Conclusion
After detailed analysis, we have identified an ideal point
of change to be taken into consideration while examining the
behaviour of an investor, following heuristic approach in
investment planning. We have observed that in between the
said threshold points the decision of the optimistic investor is
very close to the decision of the optimum investor. Before the
first threshold point and beyond the second threshold point the
behaviour of the optimistic investor and the pessimistic investor
is same as the behaviour of the optimum investor. Thus, it can
be claimed that heuristically one can reach the optimum
solution. For those investors who find it difficult to
mathematically/statistically arrive at sophisticated optimum
solution, they can heuristically obtain a portfolio similar to the
optimum portfolio. The value system of the investor under
consideration has been numerically demonstrated given the
minimum return constraints. The heuristic method is very simple
to implement. However, a heuristic solution can be at best at par
with the optimum solution but can never be better than the
optimum solution.
V{tÑàxÜ G
fxÄxvà|ÉÇ Éy cÉÜàyÉÄ|É Utáxw ÉÇ VÉxyy|v|xÇà Éy bÑà|Å|áÅ
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
100 | P a g e
4.1 Introduction
In asset allocation analysis one mostly uses the mean –
variance approach for analyzing the trade off between risk and
expected return. Since the seminal work of Markowitz (1952,
1959) the computational aspects of finding efficient portfolios
have been a concern of the finance profession. In Markowitz
portfolio theory, the portfolio of financial assets is characterized
by a desired property, the ‘reward’, and something undesirable,
the ‘risk’. Balancing financial aspects against statistical and,
particularly, computational limitations, Markowitz identified
reward and risk with the expectation and the variance of returns
respectively.
Since 1952, many works have been done to select an
optimum portfolio. But in most of the cases it is difficult to arrive
at an optimal solution. The main problem of selection of
optimum portfolio is to obtain correct and precise portfolio
weights (the proportion of wealth invested in each individual
asset). If weights can be precisely estimated, it can reduce
unnecessary transaction costs and help to construct the
optimum portfolio. From a theoretical point of view, one should
optimize expected utility for which the mean-variance approach
can be an option. Some other techniques are there to optimize
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
101 | P a g e
the portfolio. Heuristically one can reach to the optimum or near
optimum solution. At a given point of time and a given set of
conditions, heuristic solution may give reasonable result.
The portfolio optimization problem is mainly concerned
with selecting the optimal investment strategy of an investor
without considering the value system. In other words, the
investor looks for an optimal decision on how many shares of
which security she should purchase to maximize her expected
utility. If the investor knows the securities that may give her
maximum expected return or minimum expected risk, it is easy
for her to take optimal decision. But in real world it is difficult to
find out those securities due to presence of efficient market. In
this thesis we have tried to find out a heuristic solution with four
different value systems which can help an investor to take
decision which is near to the optimum decision. In this thesis we
have considered the choice of the securities, construction of
distance metric and value system of the investor.
4.2 Earlier works
Li, Yang and Deng (2007) selected optimal portfolio with
earning – at – risk (EaR). They used EaR of terminal wealth as
a measure of risk. They obtained closed form expressions for
the best constantly – rebalanced portfolio investment portfolio
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
102 | P a g e
strategy and the efficient frontier of the mean – EaR analysis to
the classical mean – variance analysis and to the mean – CaR
(capital - at - risk) analysis.
Doganoglu, Hartz and Mittnik (2007) proposed a practical
approach to portfolio selection which considers conditionally
varying volatility and fat – tailedness of risk factors explicitly into
account, while retaining analytical tractability and ease of
implementation. They used nine German DAX stocks to
illustrate their model and they found that the model is strongly
favoured by the data and it is practically implemental.
Aydin Ulucan (2007) investigated optimal holding period
for the classical mean-variance portfolio optimization problem.
He took 648 cases of Istanbul Stock Exchange ISE-100 index
and Athens Stock Exchange FTSE-40 index data. He used
three different target return levels, three different risk measures
and 36 different time periods with rolling data. He found that
portfolio returns with varying holding periods have a convex
structure with an optimal holding period.
Golosnoy and Okhrin (2007) published a paper which
proposes a multivariate shrinkage estimator for the optimal
portfolio weights. The estimated classical Markowitz weights are
shrunk to the deterministic target portfolio weights. To derive the
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
103 | P a g e
explicit solutions for the optimal shrinkage factors, they
assumed that log asset returns to be i.i.d. Gaussian. They also
empirically studied the competing portfolio selection
approaches. They prescribed that both simulation and empirical
studies show the robustness of the shrinkage estimator and that
provide significant gains to the investor compared to benchmark
procedures.
Ma and Pohlman (2008) discussed about the general
interpretation of quantile regression in the financial market. After
exploring the distributional impact of factors on returns of
securities they have found that factor effects on return depends
on the quantiles of return. They have proposed two methods for
the return forecasting and portfolio construction. Their models
show that under mild conditions more accurate forecasts can be
got and one can get more value added portfolios than the
classical conditional mean method.
Danielsson, Jorgenson, Vries and Yang (2008) proposed
optimal portfolio allocation subject to a budget constraint and a
probabilistic VaR (value – at - risk) constraint in complete
markets environments with a finite number of states. Their
method restored monotonicity of the optimal portfolio allocation
in the state-price density and reduced computational complexity.
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
104 | P a g e
Huang (2008) discussed portfolio selection problem in
combined uncertain environment of randomness and fuzziness
in his paper. In this paper, security returns were regarded as
random fuzzy variables, i.e. random returns with fuzzy expected
values. He proposed new mean – variance model and a hybrid
intelligent algorithm to solve the new model problem. He
reached his objectives with the empirical analysis.
Shaw, Liu and Kopman (2008) studied a portfolio
selection problem subject to a cardinality constraint. They
developed a dedicated Lagrangian relaxation method to
optimize the portfolio. They empirically tested the algorithm to
S&P 500 index, S&P 100 index, FTSE 100 index and FTSE 250
index.
Markus and Thorsten (2008) suggested that a time –
varying estimation of return correlations fits the data
considerably better than time – invariant estimation and thus,
increases the efficiency of risk estimation and portfolio selection.
According to them for the selection of optimal portfolios and for
risk assessment the dynamic and distributional behaviour of risk
factors are very vital.
We have proposed in chapter 3 heuristic solutions to
reach near the optimum portfolio, considering two extreme value
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
105 | P a g e
systems of investors covering optimistic and pessimistic
investors. We have examined closeness between these
heuristic solutions and the optimum solution with the help of a
distance measure. Euclidean Distance technique has been used
by them to show the similarity between the optimum portfolio
and the heuristic portfolios. For the purpose of empirical
analysis, we have selected ten companies of Nifty and the data
range they have taken from 2006 to 2007.
4.3 Optimization Framework
Following the existing literature, there are two ways
through which a portfolio can be optimally constructed. One
method is to minimize risk subject to a minimum return
constraint. The other way is to maximize the expected return
subject to a certain level of risk. We propose to strike a balance
between these two approaches by allowing either the minimum
return or the maximum risk to vary. This variation is to be
incorporated in the mathematical programming framework in
terms of coefficient of optimism (α) in the constraint set. The
optimization problem considered for the current work is of the
following type:
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
106 | P a g e
Min ````ww
Subject to minmax````
)1( Rw
Depending on the choice of we get an optimum solution
of this problem. Thus, weight vector will be a function of . One
can get the optimum portfolio corresponding to the minimum
return of the security by considering = 0 and the optimum
portfolio corresponding to the maximum return of the security by
considering =1. However, for the purpose of empirical study,
we have varied from 0 to 1 with step 0.125 resulting in nine
steps. Here, = 0 represents pessimistic investors and = 1
represents optimistic investors. For random selector and risk
planner’s the value of lies between 0 and 1. In each case, one
can get different optimum weights and with those weights return
and risk of the optimum portfolio can be calculated using
equations (3) and (4) as shown in chapter 3.
4.4 Heuristic Framework
In the next phase we have developed a heuristic solution
with the help of which we may get an alternative choice of
portfolio based on propensity to take risk. In fact, we have
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
107 | P a g e
constructed four types of heuristic portfolios covering optimistic
investor, pessimistic investor, risk planner and random selector.
In case of heuristic optimistic portfolio and heuristic
pessimistic portfolio the formulae of weights, risk and return
developed in chapter 3 is used here. In case the investor is a
risk planner, we have assumed that he/she will give importance
on both risk and return. Here we have considered both risk and
return to calculate the weights of the securities. As risk planner
investor considers both return and risk to select portfolio.
For risk planner following formula has been used for
calculating the weights of the securities of the portfolio where
both return and risk have been considered. Basically this is a
combination of the two extreme situations mentioned earlier.
)(
)( w
HRP
i
i
i
i
i
i
i
i
i
Min
Min
(11)
Given this weight function we have calculated the expected
return and risk as follows:
n
i i
i
i
i
n
ii
i
i
i
i
HRPP
R
R
1
1
)min(
)min(
(12)
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
108 | P a g e
n
i
n
jijji
j
j
j
j
i
i
i
in
i i
i
i
i
HRPP r
1 1
1
)min)(min()min(
1
(13)
For random selector, we have considered equal weight to
each security as they do not put emphasis either on return or on
risk. The weight for random selector is 1/n. Given this equal
weights we have calculated the expected return and risk for the
random selector as follows:
n
ii
HRSp R
nR
1
1 (14)
n
i
n
jijji
HRSp r
n 1 1
1 (15)
4.5 Closeness between Heuristic and Optimization framework
The closeness between the optimum framework and heuristic
framework is measured by applying the City Block Distance. The
closeness is observed to know whether an investor heuristically
can reach to near optimum solution.
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
109 | P a g e
4.6 Data Analysis and Interpretation
Average return and risk for the securities so selected
have been calculated with the help of equation 1 and 2
respectively to arrive at the objective. Classical optimum
technique has been used to optimize the portfolio. Optimum
weights of the ten securities have been calculated under
different minimum return constraints. The results are shown
below:
Table 4.1: Table showing optimum weights of fifteen
securities under different values of coefficient of optimism
Securities α values
0 0.125 0.250 0.375 0.500 0.625 0.750 0.875 1
JINDALSTEL
0 0 0 0 0 0 0 0.04870561
0
RANBAXY
0.1936191
0.1936190
0.1936186
0.1936176
0.1936158
0.1935900
0.1936190
0.07933459
0
RELIANCE
0.06280846
0.06280902
0.06281166
0.06281766
0.06282900
0.06289579
0.06280901
0.07960539
0
HDFC
0.1424088
0.1424080
0.1424066
0.1424035
0.1423969
0.1423651
0.1424096
0.4121370
1
TATAMOTORS
0.005869277
0.005868693
0.005866985
0.005863447
0.005856850
0.005814507
0.005869393
0 0
HINDALCO
0.05411821
0.05411855
0.5411938
0.05412130
0.05412621
0.05418588
0.05411801
0.01009264
0
WIPRO
0 0 0 0 0 0 0 0 0
AMBUJACEM 0.1068 0.1068 0.1068 0.1068 0.1068 0.1068 0.1068 0 0
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
110 | P a g e
151 155 160 164 158 162 148
SIEMENS
0.07751241
0.07751266
0.07751307
0.07751375
0.07751504
0.07754704
0.07751214
0.05838301
0
TATASTEEL
0 0 0 0 0 0 0 0 0
ACC
0.02198061
0.02197995
0.02197871
0.02197696
0.02197645
0.02202873
0.02198111
0 0
RELINFRA
0 0 0 0 0 0 0 0 0
AXISBANK
0.4030778
0.04030851
0.04030944
0.04031073
0.04031108
0.04028933
0.04030694
0 0
BPCL
0.1195407
0.1195405
0.1195397
0.1195376
0.1195332
0.1195004
0.1195407
0.06761105
0
SUNPHARMA
0.1750195
0.1750197
0.1750200
0.1750210
0.1750237
0.1749671
0.1750193
0.2441307
0
The weights of optimistic, pessimistic and risk planner solutions
have been calculated using the equations (5), (8) and (11)
respectively. Table 4.2 shows the weights of the optimistic,
pessimistic, risk planner and random selector solutions.
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
111 | P a g e
Table 4.2: Weights of heuristic optimistic, heuristic
pessimistic, heuristic risk planner and random selector
solutions
Securities Heuristic Optimistic Weights
Heuristic Pessimistic
Weights
Heuristic Risk
Planner Weights
Heuristic Random Selector Weights
JINDALSTEL
0.090668 0.053313 0.096953 0.066667
RANBAXY
0.066351 0.074502 0.06138 0.066667
RELIANCE
0.077863 0.076569 0.077593 0.066667
HDFC
0.101525 0.075374 0.112735 0.066667
TATAMOTORS
0.059956 0.065455 0.057583 0.066667
HINDALCO
0.072906 0.065483 0.074128 0.066667
WIPRO
0.064353 0.059075 0.066541 0.066667
AMBUJACEM
0.051962 0.077734 0.03797 0.066667
SIEMENS
0.073386 0.067679 0.07398 0.066667
TATASTEEL
0.056439 0.064709 0.053592 0.066667
ACC
0.061715 0.071019 0.056634 0.066667
RELINFRA
0.064433 0.059468 0.066428 0.066667
AXISBANK
0 0.054746 0 0.066667
BPCL
0.071308 0.064254 0.072559 0.066667
SUNPHARMA
0.087136 0.07062 0.091924 0.066667
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
112 | P a g e
Our objective is to examine the closeness between the
optimum solution and heuristic solutions. For this purpose firstly,
we have calculated the expected return and risk of the optimum
portfolio by using equation 3 and 4 as shown in chapter 3. To
calculate these expected return and risk we have used the
weights shown in Table 4.1. These expected return and risks of
the optimum portfolio under different values of the coefficient of
optimism (α) are shown in Table 4.3.
Table 4.3: Expected return and risk of the optimum portfolio
under different α values
Values of α
Minimum Return
Constraints
Expected Returns
Expected Risk
0 -0.0057 -0.0032105 0.024384
0.125 -0.00490625 -0.00114 0.015233
0.250 -0.0041125 -0.00169801 0.025905757
0.375 -0.00331875 -0.001142747 0.015233
0.500 -0.002525 -0.0011427 0.015233
0.625 -0.00173125 -0.00114272 0.015232856
0.750 -0.0009375 -0.00114273 0.0152332
0.875 -0.00014375 -0.00014375 0.017076738
1 0.00065 0.00065 0.027169
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
113 | P a g e
To study the closeness between the optimum choice of portfolio
and the heuristic choice of portfolio, we have next calculated
expected return and risk of the heuristic portfolios using
equations 6 & 7, 9 & 10 shown in chapter 3, and equations 12 &
13 and 14 & 15. Weights of the heuristic portfolios shown in
Table 4.2 are used to calculate these expected return and risk.
Table 4.4 shows the values of expected returns and risks of the
heuristic portfolios.
Table 4.4: Values of expected returns and risks of heuristic
portfolios
Heuristic portfolios Expected Return Expected Risk
Optimistic -0.001074532 0.016490497
Pessimistic -0.001478186 0.016463641
Risk Planner -0.001016865 0.016534909
Random Selector -0.001530278 0.016738164
However, for a formal look at the behaviour of heuristic
solutions in respect of the optimum solutions, we would like to
take into consideration a standard measure of distance. For this
purpose City Block Distance has been considered. With the
expected returns and risks of the optimum solution and heuristic
solutions we have computed City Block Distances (CBD) among
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
114 | P a g e
the optimum portfolio and the heuristic portfolios. With that City
Block Distance we want to know the similarity between the
optimum choice of portfolios and the heuristic choice of
portfolios. For this purpose, we have graphically shown the City
Block Distances between the optimum choice of portfolio and
the heuristic choices of portfolios. Given four heuristic solutions
we can pair wise compare them resulting in six combinations.
We have drawn six graphs. Each graph shows the City Block
Distance between the optimum solution and two heuristic
solutions. In the horizontal axis we have measured the values of
minimum return constraints and in the vertical axis we have
measured the values of City Block Distance. Graphs are shown
below:
Table 4.5: City Block Distance between the optimistic &
optimum choices of portfolios and pessimistic & optimum
choices of portfolios
Values of α Minimum Return
Constraints
Optimistic Pessimistic
0 -0.0057 0.001325534 0.001565927
0.125 -0.00490625 0.001325535 0.00156592
0.250 -0.0041125 0.001325538 0.001565915
0.375 -0.00331875 0.001325548 0.001565917
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
115 | P a g e
0.500 -0.002525 0.001325556 0.001565924
0.625 -0.00173125 0.001325831 0.001566249
0.750 -0.0009375 0.001417119 0.001793917
0.875 -0.00014375 0.001517023 0.001947533
1 0.00065 0.012403035 0.012833545
Figure 4.1
In Figure 4.1, there is one prominent threshold point T1
where the value of α is 0.875. Before the point T1 the optimistic
choice of portfolio placed near to the optimum portfolio than that
of the pessimistic choice of portfolio. Beyond the threshold point
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
116 | P a g e
T1, the two CBD curves overlapped each other. That means
beyond the point T1 the decisions of both the optimistic and
pessimistic investor are same. Thus, it can be stated that from 0
to 0.875 value of α (coefficient of optimism), the optimistic
investor’s decision is near to the optimum decision as compared
to the pessimistic decision and from 0.875 to 1 value of α the
optimistic and pessimistic investors’ decisions are same.
Table 4.6: City Block Distance between the optimistic &
optimum choices of portfolios and risk planner & optimum
choices of portfolios
Values of α Minimum Return
Constraints
Optimistic Risk Planner
0 -0.0057 0.001325534 0.001427613
0.125 -0.00490625 0.001325535 0.001427614
0.250 -0.0041125 0.001325538 0.001427617
0.375 -0.00331875 0.001325548 0.001427627
0.500 -0.002525 0.001325556 0.001427635
0.625 -0.00173125 0.001325831 0.00142791
0.750 -0.0009375 0.001417119 0.001403864
0.875 -0.00014375 0.001517023 0.001414944
1 0.00065 0.012403035 0.012300956
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
117 | P a g e
Figure 4.2
In Figure 4.2, T2 and T3 are the threshold points. To the
left of the threshold point T2 where the value of α is 0.736, the
optimistic investor’s decision is closer than that of the risk
planner decision, though the distances are very less. It means
that the optimistic investor’s solution is closer to the optimum
solution than that of the risk planner’s solution. In between T2
and T3 (value of coefficient of optimism α is 0.875) the risk
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
118 | P a g e
planner’s decision is closer to the optimum decision than that of
the optimistic investor’s decision. On the other hand, in the right
side of the point T3, the CBD between optimum and optimistic
portfolios is overlapping with the CBD between optimum and risk
planner portfolios. That means the decisions of the optimistic
and the risk planner is same beyond 0.875 value of α.
Table 4.7: City Block Distance between the optimistic &
optimum choices of portfolios and random selector &
optimum choices of portfolios
Values of α Minimum Return
Constraints
Optimistic Random Selector
0 -0.0057 0.001325534 0.001892542
0.125 -0.00490625 0.001325535 0.001892535
0.250 -0.0041125 0.001325538 0.00189253
0.375 -0.00331875 0.001325548 0.001892532
0.500 -0.002525 0.001325556 0.001892539
0.625 -0.00173125 0.001325831 0.001892864
0.750 -0.0009375 0.001417119 0.002120532
0.875 -0.00014375 0.001517023 0.001725102
1 0.00065 0.012403035 0.012611114
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
119 | P a g e
Figure 4.3
From the above Figure, it is observed that to the left of
the threshold point T4, where the value of coefficient of optimism
is 0.757, optimistic investor’s decision is closer to the optimum
solution than that of random selector’s decision. In other words,
from 0 to 0.757 values of coefficient of optimism the optimistic
investor’s decision is better than that of random selector’s
decision. But beyond the point T4 (or from the value of α 0.757
to 1) both optimistic and random selector shows the same
decision.
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
120 | P a g e
Table 4.8: City Block Distance between the pessimistic &
optimum choices of portfolios and risk planner & optimum
choices of portfolios
Values of α Minimum Return
Constraints
Pessimistic Risk Planner
0 -0.0057 0.001565927 0.001427613
0.125 -0.00490625 0.00156592 0.001427614
0.250 -0.0041125 0.001565915 0.001427617
0.375 -0.00331875 0.001565917 0.001427627
0.500 -0.002525 0.001565924 0.001427635
0.625 -0.00173125 0.001566249 0.00142791
0.750 -0.0009375 0.001793917 0.001403864
0.875 -0.00014375 0.001947533 0.001414944
1 0.00065 0.012833545 0.012300956
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
121 | P a g e
Figure 4.4
From the above Figure 4.4 it can be stated that the CBD
between the optimum portfolio and the risk planner’s portfolio
gives better result than the CBD between the optimum portfolio
and the pessimistic portfolio. What is important to note is that no
threshold point can be identified. The risk planner gives better
decision than that of the pessimistic investor through out the
whole range of coefficient of optimism.
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
122 | P a g e
Table 4.9: City Block Distance between the pessimistic &
optimum choices of portfolios and random selector &
optimum choices of portfolios
Values of α Minimum Return
Constraints
Pessimistic Random Selector
0 -0.0057 0.001565927 0.001892542
0.125 -0.00490625 0.00156592 0.001892535
0.250 -0.0041125 0.001565915 0.00189253
0.375 -0.00331875 0.001565917 0.001892532
0.500 -0.002525 0.001565924 0.001892539
0.625 -0.00173125 0.001566249 0.001892864
0.750 -0.0009375 0.001793917 0.002120532
0.875 -0.00014375 0.001947533 0.001725102
1 0.00065 0.012833545 0.012611114
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
123 | P a g e
Figure 4.5
When pessimistic investor and random selector are
chosen for comparison it is observed that the decision of both
the pessimistic investor and random selector is very close to the
optimum solution beyond the threshold point T6 (value of α is
0.875). Between the points T5 (value of α 0.824) and T6 random
selector’s decision is closer to the optimum solution than that of
the pessimistic investor’s decision. But before the threshold
point T5 the pessimistic investor’s decision gives better result as
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
124 | P a g e
compared to the random selector’s decision, as the CBD
between the optimum portfolio and the pessimistic portfolio is
closer than that of the optimum portfolio and random selector’s
portfolio. Thus, it can be stated that in between 0<α<0.824 the
pessimistic solution is closer to the optimum solution than that of
the random selector’s solution and in between 0.824 and 0.875
the random selector shows the better result as compared to
pessimistic investor. Beyond the value of α 0.875 both heuristic
portfolios show same result.
Table 4.10: City Block Distance between the pessimistic &
optimum choices of portfolios and random selector &
optimum choices of portfolios
Values of α Minimum Return
Constraints
Risk Planner Random Selector
0 -0.0057 0.001427613 0.001892542
0.125 -0.00490625 0.001427614 0.001892535
0.250 -0.0041125 0.001427617 0.00189253
0.375 -0.00331875 0.001427627 0.001892532
0.500 -0.002525 0.001427635 0.001892539
0.625 -0.00173125 0.00142791 0.001892864
0.750 -0.0009375 0.001403864 0.002120532
0.875 -0.00014375 0.001414944 0.001725102
1 0.00065 0.012300956 0.012611114
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
125 | P a g e
Figure 4.6
By putting the CBD values related to the risk planner and
random selector, T7 is marked as the threshold point where the
value of α is 0.97 in Figure 4.6. When the decision between risk
planner and random selector is compared from the above figure,
it is found that to the left hand side of the threshold point T7 the
risk planner’s solution is giving better result than that of the
random selector’s solution. But beyond point T7 (or the value of
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
126 | P a g e
α 0.97) the risk planner’s decision is same as the random
selector’s decision, because the CBD between the optimum
portfolio and the risk planner’s portfolio is same as the CBD
between the optimum portfolio and random selector’s portfolio.
After completing this pair-wise comparison, a combined
graph is drawn which shows all the six combinations so that the
distances of all the heuristic solutions with the optimum solution
can be examined together. Like the above graphs in Figure 4.7,
on the horizontal axis minimum return constraints are measured
and on the vertical axis CBD values are measured.
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
127 | P a g e
Table 4.11: City Block Distance between optimistic and
optimum solutions, between pessimistic and optimum
solutions, between risk planner and optimum solutions and
between random selector and optimum solutions
Values of α
Minimum Return
Constraints
Optimistic Pessimistic Risk Planner
Random Selector
0 -0.0057 0.001325534
0.001565927
0.001427613
0.001892542
0.125 -0.00490625 0.001325535
0.00156592
0.001427614
0.001892535
0.250 -0.0041125 0.001325538
0.001565915
0.001427617
0.00189253
0.375 -0.00331875 0.001325548
0.001565917
0.001427627
0.001892532
0.500 -0.002525 0.001325556
0.001565924
0.001427635
0.001892539
0.625 -0.00173125 0.001325831
0.001566249
0.00142791
0.001892864
0.750 -0.0009375 0.001417119
0.001793917
0.001403864
0.002120532
0.875 -0.00014375 0.001517023
0.001947533
0.001414944
0.001725102
1 0.00065 0.012403035
0.012833545
0.012300956
0.012611114
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
128 | P a g e
Figure 4.7
In the above Figure 4.7 it is observed that before the
threshold point T8 (where the value of coefficient of optimism α
is 0.736) optimistic solution is giving better result than that of all
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
129 | P a g e
other heuristic solutions. Then the risk planner’s solution is
showing better result than that of the pessimistic solution and
random selector’s solution. The random selector’s solution is
worst before the point T8. Between the threshold point T8 and
T10 (value of α is 0.875) the risk planner’s decision is better
than the optimistic investor’s decision, as the CBD between the
optimum portfolio and optimistic portfolio is situated far than the
CBD between the optimum portfolio and the risk planner’s
portfolio. At the threshold point T9 where the value of α is 0.824
a little change is occurred. Beyond the point T9 the random
selector’s decision is placing in the third position irrespective of
the pessimistic investor. In between the points T9 and T11 the
decision of the pessimistic investor is worst. But beyond the
threshold point T11 all the heuristic solutions are showing same
result. In other words, it can be stated that beyond the threshold
point T11 the CBD between the optimum portfolio and the four
heuristic portfolios are overlapping. Thus, beyond the point T11
all the heuristic portfolios give the same decision.
Chapter 4: Selection of Portfolio based on Coefficient of Optimism
130 | P a g e
4.7 Conclusion
After detailed analysis, a mixed result has been obtained.
When the value of coefficient of optimism lies between 0 and
0.750 the decision of the optimistic investor is best than the
other heuristic investors though the risk planner’s decision is
very close to the optimistic investor’s decision. Only at 0.750
and 0.850 values of coefficient of optimism the risk planner’s
decision is best as compared to the other heuristic solutions. But
at the coefficient of optimism 1, all the four heuristic solutions
show the same result. 0=<α>=0.824 the random selector shows
the worst result. In this range the pessimistic investor is having
the third best position. Thus, we can say that heuristically the
optimum solution can be reached. In other words, it can be said
that closeness is there between the optimum portfolio
constructed on the basis of existing techniques and portfolio
selected by investors who follow heuristic approach.
V{tÑàxÜ H
itÄâx Utáxw fxÄxvà|ÉÇ Éy cÉÜàyÉÄ|ÉM T [xâÜ|áà|v
TÑÑÜÉtv{
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
131 | P a g e
5.1 Introduction
The portfolio optimization problem aims to find out the
optimal investment strategy of an investor. In other words, the
investor looks for an optimal decision on how many shares of
which security he/she should invest to maximize his/her
expected return or minimize his/her risk. Markowitz (1959) was
the first person who had shown that through mean – variance
analysis an investor can get optimum portfolio. According to
Markowitz model, we need large information to calculate an
efficient portfolio. Sharpe (1964) worked on the same problem
and developed Capital Asset Pricing Model (CAPM) which can
be used for multiple securities to construct an efficient portfolio
and can be used with relatively less amount of information.
Inspired by their seminal works many researchers have been
doing work on optimum choice of the portfolio. Amongst them
Merton (1969, 1971, 1973, and 1980) is the lead researcher who
discussed the continuous time portfolio optimization problem in
the utility framework. Brown (1978) studied the problem when an
investor has only two options - a single risky asset and an asset
with a constant and known rate of return. He considered Optimal
Baye’s Portfolio and Certainty Equivalence Strategy to calculate
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
132 | P a g e
the efficiency of the asset. He found that the optimal Baye’s
portfolio invests less in risky securities.
5.2 Earlier works
While addressing the issue whether the systematic risk
plays any role to choose a portfolio of international equities, Das
and Uppal (2004) developed two models- one that incorporates
systemic risk and the other that ignores it. The models of asset
returns help to capture the jumps, which reflects the returns on
international equities, occurring at the same time across
countries. They also showed how an investor would choose an
optimal portfolio when returns have these features. They tested
their method to determine the weights for a portfolio allocated
over a risk less asset. However, they have got the result that “for
investors with low risk aversion who desire levered positions, the
cost of ignoring systemic risk is larger, and in the case of a
highly levered portfolio, there is a positive probability of losing
one’s entire wealth if there is a large negative systemic shock”.
So to analyse a portfolio, the systematic risk plays an important
role.
On the other hand, Jones (1999) showed that for
empirical analysis of classical mean-variance problem,
regression method is a simple tool to portfolio analysis. By using
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
133 | P a g e
GRS F- test (formulated by Gibbons, Ross, and Shanken
(1989)) of portfolio efficiency he showed that a test of the
restriction, the weights of the ‘tangency’ portfolio equal the
weights of the test portfolio. He also proposed the
implementation of GRS F-test for portfolio efficiency using linear
restrictions on a single linear OLS regression. He analyzed the
weights of an international efficient portfolio. He used monthly
data from Morgan Stanley Capital International (MSCI) for the
20-years for the equity markets of 11 developed countries.
Thus, over the years, many authors have used various
techniques to find out an optimum portfolio that can help an
investor to take the optimum decision. Some of them gave
emphasis on the number of assets in the portfolio; others
assigned importance on the weights of the assets of the
portfolio. Recently works are on the suitability of heuristic
approach for solving this complex problem. In this thesis we
have considered two types of portfolios – optimum portfolio and
heuristic portfolio and determined weight of the portfolio, i.e. rate
of investment in each asset of the portfolio. Calculation of
assets’ weight is the most crucial task of the investors. If
investor can correctly predict the proportion of amount to be
invested in each assets of the portfolio, he/she can take right
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
134 | P a g e
decision to earn maximum utility. But in real world in an efficient
market, it is very difficult to predict assets’ weights accurately.
The problem of arriving at weights of the assets in the portfolio
has been discussed from various angles by different authors.
Though optimum portfolio is conceptually the best choice
yet investors opt for conditional portfolios given their own value
systems. This calls for detailed study on effects on value
systems on the optimum choice of the portfolio. In this thesis we
have tried to examine those effects. In chapter 3 we have
considered two human value systems in terms of propensity to
take risk. These are optimistic investor and pessimistic investor.
But these are two extreme states of an investor’s value. In this
chapter we would like to generalize this concept in terms of
dependency on coefficient of optimism of the investor. We would
like to compare this heuristic formulation with the optimum
portfolio.
5.3 Choice of weights under different value systems
In the next stage, we propose a heuristic solution to get
an alternative choice of portfolio based on the same propensity
to take risk. For this purpose, we have considered three types of
investor viz. optimistic investor, pessimistic investor and risk
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
135 | P a g e
planner. Since in chapter 3 we have considered optimistic
portfolio and pessimistic portfolio and calculated their closeness
from the optimum portfolio we propose to strike a balance
between these two polar opposite states to describe all possible
situations and present a general heuristic portfolio with α as the
coefficient of optimism.
n
i
HPi
HOi
HPi
HOi
H
ww
www
1
)2/1( (16)
We would like to observe the closeness between the
optimum portfolios and the heuristic portfolios with an effect of
coefficient of optimism. Whether coefficient of optimism has any
effect on choice of a heuristic portfolio? To find out the answer
of the question, we have considered the following weight
generated with the help of equation 17. We assume that a
general investor gives emphasis both on the return and risk.
Weights of optimistic investor and pessimistic investor have
been considered together with the value of coefficient of
optimism to calculate the weight of the heuristic portfolio. The
formula is given below:
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
136 | P a g e
n
i
HPi
HOi
HPi
HOi
H
ww
www
1
)1(
)1(
)(
(17)
Given the above weight function one may calculate the
expected return and risk of the general heuristic portfolio with
the following formula.
n
i
n
j
HPi
HOi
n
ii
HPi
HOi
H
ww
RwwR
1 1
1
1
1
(18)
n
i
n
jijji
HPj
HOj
HPi
HOi
HPi
HOi
H rwwwwww 1 1
11
1
))((1
(19)
It is not possible to analytically examine the closeness of
the heuristic portfolio with the optimum portfolio; empirical
analysis is the only way out to undertake a comparative study
5.4 Data Analysis and Interpretation
Average return and risk for the securities so selected
have been calculated by applying equations 1 and 2 shown in
chapter 3. For nine different values of α coefficient optimum
weights of the fifteen securities have been calculated. These
optimum weights are given in Table 5.1:
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
137 | P a g e
Table 5.1: Weights of the optimum portfolios of fifteen
selected securities under different α coefficients
Securities
α values
0 0.125 0.250 0.375 0.500 0.625 0.750 0.875
1.000
JINDALSTEL
0 0 0 0 0 0 0 0.048705
61
0
RANBAXY
0.1936191
0.1936190
0.1936186
0.1936176
0.1936158
0.1935900
0.1936190
0.079334
59
0
RELIANCE
0.0628084
6
0.0628090
2
0.0628116
6
0.0628176
6
0.0628290
0
0.0628957
9
0.0628090
1
0.079605
39
0
HDFC
0.1424088
0.1424080
0.1424066
0.1424035
0.1423969
0.1423651
0.1424096
0.412137
0
1
TATAMOTO
RS
0.0058692
77
0.0058686
93
0.0058669
85
0.0058634
47
0.0058568
50
0.0058145
07
0.0058693
93
0 0
HINDALCO
0.0541182
1
0.0541185
5
0.5411938
0.0541213
0
0.0541262
1
0.0541858
8
0.0541180
1
0.010092
64
0
WIPRO
0 0 0 0 0 0 0 0 0
AMBUJACE
M
0.1068151
0.1068155
0.1068160
0.1068164
0.1068158
0.1068162
0.1068148
0 0
SIEMENS
0.0775124
1
0.0775126
6
0.0775130
7
0.0775137
5
0.0775150
4
0.0775470
4
0.0775121
4
0.058383
01
0
TATASTEEL
0 0 0 0 0 0 0 0 0
ACC 0.0219806
1
0.0219799
5
0.0219787
1
0.0219769
6
0.0219764
5
0.0220287
3
0.0219811
1
0 0
RELINFRA
0 0 0 0 0 0 0 0 0
AXISBANK
0.4030778
0.0403085
0.0403094
0.0403107
0.0403110
0.0402893
0.0403069
0 0
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
138 | P a g e
1 4 3 8 3 4
BPCL 0.1195407
0.1195405
0.1195397
0.1195376
0.1195332
0.1195004
0.1195407
0.067611
05
0
SUNPHARM
A
0.1750195
0.1750197
0.1750200
0.1750210
0.1750237
0.1749671
0.1750193
0.244130
7
0
These optimum weights under different α coefficients
have been used to calculate the expected return and risk of the
optimum portfolios. In this manner we have nine optimum
portfolios to be considered. Equations 3 and 4 as shown in
chapter 3 have been used to calculate the expected risk and
return of the respective optimum portfolios. In the following table
the expected return and risk of the optimum securities under
different values of coefficient of optimism are given in Table 5.2:
Table 5.2: Expected return and risk of the fifteen selected securities under different α coefficients
α coefficient Expected Return Expected Risk
0 -0.0032105 0.024384
0.125 -0.00114 0.015233
0.250 -0.00169801 0.025905757
0.375 -0.001142747 0.015233
0.500 -0.0011427 0.015233
0.625 -0.00114272 0.015232856
0.750 -0.00114273 0.0152332
0.875 -0.00014375 0.017076738
1.000 0.00065 0.027169
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
139 | P a g e
In the next phase we have considered the two heuristic
portfolios for optimistic investor and pessimistic investor. The
weights of the heuristic portfolios have been calculated by using
equations 5 and 8 respectively. Table 5.3 represents the weights
of these two heuristic portfolios.
Table 5.3: Weights of two heuristic portfolios taken into
consideration
Securities Optimistic Portfolio Pessimistic Portfolio
JINDALSTEL 0.090668 0.053313
RANBAXY 0.066351 0.074502
RELIANCE 0.077863 0.076569
HDFC 0.101525 0.075374
TATAMOTORS 0.059956 0.065455
HINDALCO 0.072906 0.065483
WIPRO 0.064353 0.059075
AMBUJACEM 0.051962 0.077734
SIEMENS 0.073386 0.067679
TATASTEEL 0.056439 0.064709
ACC 0.061715 0.071019
RELINFRA 0.064433 0.059468
AXISBANK 0 0.054746
BPCL 0.071308 0.064254
SUNPHARMA 0.087136 0.07062
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
140 | P a g e
After calculating the weights of the optimistic and
pessimistic investors we have calculated the weights of the
selected securities of the general heuristic portfolio. For this
purpose equation 17 has been used. The weights of the
securities of the general heuristic portfolios under different
coefficient of optimism values are given in Table 5.4:
Table 5.4: Weights of the heuristic portfolios of fifteen selected securities under different α coefficients
Securities
α values
0 0.125
0.250
0.375
0.500
0.625
0.750
0.875
1.000
JINDALST
EL
0.053313
0.060000264
0.063783791
0.067757147
0.071926093
0.076296261
0.080873
13
0.085661989
0.090668
RANBAXY
0.074502
0.077333969
0.075824542
0.074291121
0.072736197
0.071162256
0.069571766
0.067967182
0.066351
RELIANC
E
0.076569
0.080808152
0.080555369
0.080245638
0.079879428
0.079457315
0.078979983
0.078448221
0.077863
HDFC
0.075374
0.082391597
0.085070973
0.087774326
0.090498248
0.093239143
0.095993226
0.098756526
0.101525
TATAMOTORS
0.065455
0.068182275
0.067086828
0.065961521
0.064808
3
0.063629136
0.062426019
0.061200949
0.059956
HINDALCO
0.065
0.069895
0.070470
0.070999
0.071480
0.071912
0.072295
0.072626
0.0729
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
141 | P a g e
48 679 803 572 702 985 296 593 06
WIPRO
0.059075
0.062884237
0.063229096
0.063530135
0.063786556
0.063997
64
0.064162
75
0.064281336
0.0643533
AMBUJAC
EM
0.077734
0.077846268
0.073637952
0.069607
05
0.065749329
0.062060441
0.058535939
0.055171303
0.051962
SIEMENS
0.067679
0.072001455
0.072354
53
0.072657059
0.072908215
0.073107267
0.073253577
0.073346609
0.073386
TATASTEE
L
0.064709
0.066993697
0.065514936
0.064022749
0.062519413
0.061007171
0.059488237
0.057964783
0.056439
ACC
0.071019
0.073492666
0.071837395
0.070168916
0.068489743
0.066802349
0.065109166
0.063412583
0.061715
RELINFRA
0.059468
0.063259956
0.063564051
0.063823683
0.064038144
0.064206801
0.064329109
0.064404611
0.064433
AXISBANK
0.054746
0 0 0 0 0 0 0 0
BPCL
0.064254
0.068556297
0.069092621
0.069583073
0.070026449
0.070421623
0.070767553
0.071063276
0.071308
SUNPHARMA
0.07062
0.076353488
0.077977116
0.079578
01
0.081153183
0.082699611
0.084214
25
0.085694
04
0.087136
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
142 | P a g e
With the weights given in Table 5.4 we have calculated
the expected return and risk of the general heuristic portfolio.
Equations 18 and 19 have been used to find out the expected
return and risk of the general heuristic portfolios respectively.
Expected return and risk of the general heuristic portfolios under
different values of coefficient of optimism is show in the following
Table 5.5.
Table 5.5: Expected return and risk of heuristic portfolios
under different α coefficients
α coefficient Expected Return Expected Risk
0 -0.00147818 0.016463595
0.125 -0.001214085 0.016412405
0.250 -0.00119441 0.016414064
0.375 -0.001174638 0.01641861
0.500 -0.001154781 0.016426197
0.625 -0.001134838 0.016436991
0.750 -0.001114814 0.016451174
0.875 -0.001094711 0.016468941
1.000 -0.001074536 0.016490524
Once we have obtained the expected return and risk of the
general heuristic portfolios under different values of coefficient of
optimism, we have considered City Block Distance to measure
the closeness of the heuristic portfolio with the optimum
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
143 | P a g e
portfolio. City Block Distance is a standard measure of distance
to calculate the similarities or dissimilarities between two points.
Since we would like to find out the similarities or dissimilarities
between the optimum portfolios and the heuristic portfolios, this
City Block Distance can be applied on a two dimensional set up
with the expected return and risk of the optimum portfolios and
the heuristic portfolios. City Block Distance between the
optimum portfolios and the heuristic portfolios under different
coefficient of optimism values are given in Table 5.6:
Table 5.6: City Block Distances between optimum portfolios
and heuristic portfolios under different α coefficients
α Coefficient City Block Distance values
0 0.009652725
0.125 0.00125349
0.250 0.009995293
0.375 0.001217501
0.500 0.001205278
0.625 0.001212017
0.750 0.00124589
0.875 0.001557958
1.000 0.012403012
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
144 | P a g e
CBD between Optimum and Heuristic Portfolios
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0 0.125 0.25 0.375 0.5 0.625 0.75 0.875 1 1.125
Values of α coefficient
CB
D v
alues
CBD between optimum and heuristic portfolios
Figure 5.1
These City Block Distances between the optimum portfolios and
heuristic portfolios are presented in a graph (see figure 5.1). In the
said graph we have presented the values of coefficient of optimism
(α coefficient) along the horizontal axis and the values of City Block
Distance along the vertical axis.
Form the above figure it is clear that when the value of
the coefficient of optimism (α coefficient) lies between 0 or 0.25
the decision of the heuristic investor does not match with the
decision of the optimum investor. It means that for 0≤ α ≤0.25,
there is less similarities between the optimum solution and the
heuristic solution. However, at the value of α at 0.125 the
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
145 | P a g e
decision of the heuristic investor is closer to the optimum
solution. The decision of the heuristic investor again does not
match when the value of α coefficient is nearly 1. For 0.375≤ α
≤0.875, the heuristic portfolio is very close to the optimum
portfolio. So, there is a marked similarity between the optimum
portfolio and the heuristic portfolio when the value of coefficient
of optimism is in between 0.375 to 0.875. Thus, we can say that
risk planners’ decision is near to optimum decision.
5.5 Conclusion
After thorough empirical analysis along with comparative
studies, it has been noted that heuristically one can get a set of
weights that provide a result very close to the optimum result.
The investor has to be aware of the coefficient of optimism while
making investment in any security. It has been observed that the
decision of the pessimistic investor does not match with the
optimum solution because when the value of coefficient of
optimism α is equal to 0 (represents the weight of the
pessimistic investor) the distance between the optimum solution
and the heuristic solution is very large. When the value of the
coefficient of optimism is 1 then also the decision of the
optimistic investor does not match with the optimum decision
because the CBD between the optimum solution and optimistic
Chapter 5: Value based Selection of Portfolio: A Heuristic Approach
146 | P a g e
solution is large. But when we have considered the value of
coefficient of optimism between 0.375 to 0.875 covering the risk
planner’s decision, the heuristic solution is very close to the
optimum decision. So the moderate risk taker may reach nearly
optimum portfolio if they considered the heuristic approach. And
mostly an investor falls in this moderate range. Thus, an investor
can easily arrive at near optimum solution heuristically.
V{tÑàxÜ I
VÉÅÑtÜtà|äx TÇtÄçá|á ÉÇ f{tÜÑxËá Vâà @Éyy cÜ|Çv|ÑtÄ tÇw t axtÜ bÑà|ÅâÅ cÜ|Çv|ÑÄx yÉÜ
cÉÜàyÉÄ|É fxÄxvà|ÉÇ âÇwxÜ `ÉwxÄ TÑÑÜÉtv{
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
147 | P a g e
6.1 Introduction
Although the pioneer work done in portfolio management
by Markowitz gained much importance in portfolio optimization
but was not without limitations. However, the next pioneering
work was done by W. Sharpe (1971) in order to overcome the
limitations of the mean-variance model of portfolio optimization.
He developed a model, also known as, Single Index Model
which describes risk and return of a portfolio with relatively less
calculations. His work was influenced by a prior work known as
Capital Asset Pricing Model (1964). In Single Index Model,
requirement of data is comparatively less than the Mean –
Variance approach. In this model Sharpe assumes that ‘the co-
movement of the stocks is due to a single common influence or
index’. That is why this model is known as Single Index Model.
Later many works have been reported in the literature on
portfolio management. Many authors have suggested selection
of portfolio in many ways. These have established a close
relationship with statistic of modeling. With various stochastic
formulations (see Marton,1980; Sahalia & Brandt, 2001;
Detemple, Garcia and Rindisscher, 2003; Beliakov and Bagirov,
2006; Okhrin and Schmid, 2008 etc.) these authors discussed
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
148 | P a g e
the problem of selection of portfolio from optimum to heuristic
point of view.
6.2 Earlier works
The relationship between the optimal portfolio obtained
by using Elton, Gruber and Padberg’s algorithm under the
Single Index Model and the optimal portfolio under the Single
Index Model with the introduction of a market security was
discussed by Stucchi (1991). The author empirically showed the
relationship and got that there is an exact relationship between
the optimal portfolio obtained by using Elton, Gruber and
Padberg’s algorithm under the Single Index Model and the
optimal portfolio under the Single Index Model with the
introduction of a market security.
Dutt (2003) constructed an optimal portfolio using
Sharpe’s Single Index Model. For analysis purpose he took data
of BSE 100 as market index from October 2002 to April 2003.
He observed that all the selected stock turn out to be bank
stocks. Antomil, J. et al (2004) discussed the application of
Fuzzy Compromise Programming to Portfolio Selection using
Sharpe’s single index model. They introduced the fuzzy ideal
solution concept based on soft preference and indifference
relationships and they canonically represented the fuzzy
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
149 | P a g e
numbers by means of their α –cuts. The fuzzy parameters are
used to evaluate the accuracy between the ideal solution and
the objective values through their expected intervals. Their
model work with data that contain more information than any
classical model and the model can easily be dealt with.
Computational effort is less in their model. Their model allows
the preference of the decision makers to select an optimal
portfolio.
Bilbao, A. et al (2006) presented Sharpe's single-index
model and on Fuzzy Sets Theory to deal with the problem of
portfolio selection. Their model has three basic concepts like
value, ambiguity and fuzziness. They have introduced ‘Expert
Betas’ which are nothing but the expert estimations about future
Betas. Each financial asset’s ‘Extra Beta’ is modelled as
trapezoidal fuzzy numbers. They have also proposed Goal
Programming model to select an optimal portfolio which can
include imprecise investor's aspirations considering both, high-
and low-risky assets. To illustrate the proposed model they have
considered a real portfolio selection problem.
Sharpe’s Single Index Model is therefore an interesting
area of research. Many authors have been done their research
work in this specific field. Chitnis (2010) constructed two
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
150 | P a g e
optimum portfolios using Sharpe’s Single Index Model and
compared them by Sharpe’s Ratio. The author used two
samples to construct the two portfolios. She took Nifty as the
market index and found that individual securities have risk return
characteristic of their own. According to her, portfolio spreads
risk over many securities and the greater the portfolio’s Sharpe’s
ratio, the better is its performance.
Mishra, A. K. (2011) studied the Sharpe’s Single Index
model to illustrate a real portfolio selection problem. For
empirical study he has considered thirty stocks listed in BSE
Sensex. He has given idea to the researcher that optimum
portfolio can be obtained through Sharp’s Single Index model in
real world.
Rani, M.,Bahl, S.(2012) used the Sharpe’s Single Index
model to construct an optimal portfolio with the procedure of
short sales and without the procedure of short sales. They
constructed an optimal portfolio and represented the optimal
portfolio with the percentage invested in each stock. They have
selected optimum portfolio with eleven stocks out of thirty stocks
in BSE Sensex. They have concluded that Sharpe’s Single
Index Model is of great importance and the framework Model for
optimal portfolio construction is very simple and useful. Any
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
151 | P a g e
investor can easily use this method to construct the optimum
portfolio. They have considered data from April 2006 to March
2011of BSE Sensex.
As discussed in chapter 4 and 5, the suggested heuristic
selection of portfolio is based on risk taking propensity. As
selection of optimum portfolio is a numerically involved task,
they examined the closeness between the heuristic portfolio and
the optimum portfolio. Basically, they have considered the
concept coefficient of optimism popular in decision theory to
calculate the optimum weights of different optimum portfolios
and studied the variation in closeness between the heuristic
portfolio and optimum portfolios with the variation in coefficient
of optimism. They noted that when the value of the coefficient of
optimism is 1 the optimistic investor’s decision is away from the
optimum decision and when the value of coefficient of optimism
is equal to 0 the pessimistic investor’s decision is not close to
the optimum decision. But when the value of coefficient of
optimism is moderate, i.e; in between 0.375 to 0.875, the
heuristic solution is very close to the optimum decision. These
mean the moderate risk taker may reach nearly optimum
portfolio if they consider the heuristic approach.
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
152 | P a g e
6.3 Sharpe’s Single Index Model
Sharpe made an attempt to predict the return on a
security based on a market return index. The basic equation
underlying the Single Index Model is
imiii eRR
The assumptions of Sharpe’s model are:
1. The expected value of ei is zero, i.e; E(ei) = 0,
2. The only reason of stock movement is a common co-
movement with the index, i.e. E(ei,ej) = 0 and
3. The index is unrelated to unique return, i.e,
0)]([ mmi RReE .
To select the securities for the Sharpe’s optimal portfolio, we
can consider those securities whose returns should be more
than the risk free return and the beta value for the securities
should be positive.
6.4Sharpe’s cut off principle
In the next phase, we have to rank the securities based
on [ ifi RR /)( ] to construct the optimal portfolio, where fR
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
153 | P a g e
Return of a risk free asset. The size of a security’s excess return
to beta ratio of [ ifi RR /)( ] decides whether a security is
included or excluded in the portfolio. Depending on a unique cut
off rate C *, all the securities which have higher ratio of
[ ifi RR /)( ] are included in the portfolio and those securities
are excluded from the portfolio which have lower ratio of
[ ifi RR /)( ].
Securities are ranked in descending order of magnitude
according to their excess return-to-beta ratio. Then, we have to
consider a cut-off rate C * to select the portfolio. The formula
developed by Sharpe to calculate the cut-off rate is given below:
n
j ei
im
n
j ej
ifim
i
RR
C
12
22
12
2
1
)(
C * is a candidate of Ci. The value of C * is that value of Ci ,
before which all the values of Ci are increasing and beyond
which all the values of Ci are decreasing. That means when the
values of Cis change its pattern we have to consider that value
as C *.
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
154 | P a g e
After selecting the securities we have to calculate the
expected return and risk of the portfolio using the following
equation 20 and 21 respectively.
n
iiip RwR
1
(20)
n
ieiim
n
iiip ww
1
222
2
1
(21)
6.5 Optimum framework under value system and Sharpe’s Single Index Model
Following the existing literature, a portfolio can be
optimized in two ways. One method is to minimize risk subject to
a minimum return constraint. The other way is to maximize the
expected return subject to a certain level of risk. We propose to
combine these two and minimize the portfolio risk subject to a
minimum return constraint. Under a given coefficient of optimism
under Sharpe’s Single Index Model, a mathematical
programming framework with minimum return constraint
expressed in terms of coefficient of optimism (α), is given below:
Min
n
imiip w
1
222 )( +
n
ieiiw
1
22 (22)
Subject to minmax1
)1()(
n
imiii Rw ,
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
155 | P a g e
As already stated, following the decision theory
framework we have allowed the values of coefficient of optimism
ranging to vary from 0 to 1 to get different sets of optimum
weights of the portfolio for different choices of the coefficient of
optimism (). These optimum solutions are different risk-based
optimum solutions obtained by raising the value of step by
step. With these optimum weights, the expected return and risk
of the optimum portfolio can be calculated. The risk and return of
the optimum portfolio for a given α have been calculated by
using the equations 20 and 21.
6.6 A Near Optimum Approach
We like to propose an alternative approach to be referred
as near optimum approach. In the proposed approach we
withdraw a term from equation 22. Since, the dropped term may
not have significant contribution we expect that our proposed
principle will offer a near optimum solution. For near optimum
portfolio we have used the following formula to reduce the
original non-linear programming problem stated at equation 22
to a linear programming problem.
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
156 | P a g e
Min iiw (23)
Subject to minmax1
1)(
n
imiii Rw
Since linear programming problem can be easily handled
one gets computational edge over the other two approaches.
We like to compare Sharpe’s cut off principle portfolio and the
proposed near optimum portfolio with that of optimum portfolio
under Sharpe’s Single Index Model and examine the suitability
of near optimum portfolio over Sharpe’s cut off principle
portfolio. The expected return and risk of the near optimum
portfolio have been calculated by applying equations 20 and 21.
6.7 Comparative study
For comparative analysis purpose 10 securities are
considered viz. 1, 2, 3, 4, 5, 6, 7, 8, 9, 10. The average returns
of the 10 securities are 15, 17, 12, 17, 11, 11, 11, 7, 7, and 5.6
respectively. The optimal portfolio consisted of 5 securities viz.
security 1, 2, 3, 4 and 5. Their Ci’s are more than the cut-off rate
which is 5.45 for the example. We have calculated the optimum
weights for nine different values of coefficient of optimism. The
optimum weights are given below in Table 6.1:
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
157 | P a g e
Table 6.1: Weights of the optimum portfolios under
Sharpe’s Single Index Model of ten securities under
different α values
Securities
α values
0 0.125
0.250
0.375
0.500
0.625
0.750
0.875
1
1 0.023774
15
0.040958
19
0.0912
8
0.132748
0.1629226
0.189429
0.215936
2
0.238697
3
0
2 0 0 0.040052
0.098201
0.1355401
0.166186
0.196831
5
0.259457
9
0.3809524
3 0.059435
36
0.083917
33
0.152155
0.205959
0.2383973
0.264528
0.290658
9
0.171594
3
0
4 0 0 0 0 0.0444953
0.103612
0.162728
5
0.315311
5
0.6190476
5 0.029717
68
0.038878
97
0.063403
0.081995
0.0910471
0.097423
0.103799
2
0.014938
98
0
6 0 0 0 0 0 0 0 0 0
7 0 0 0 0 0 0 0 0 0
8 0.161961
4
0.151647
9
0.131163
0.1048
4
0.0705187
0.035432
0.034452
56
0 0
9 0.059435
36
0.053120
39
0.025412
0 0 0 0 0 0
10 0.665676
1
0.626977
2
0.496534
0.376258
0.2570788
0.1433
9
0.029701
1
0 0
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
158 | P a g e
We have calculated expected return and risk of the optimum
portfolios under different α values. The value of expected return
and risk of the nine different optimum portfolios are show below:
Table 6.2: Expected return and risk of the optimum
portfolios under Sharpe’s Single Index Model under
different values of α
α values Minimum Return
Constraints
Expected Return
Risk
0 5.6 6.67429453 2.265304766 0.125 7.025 6.99349983 2.181982719 0.250 8.45 8.45000015 2.065744605 0.375 9.875 9.87500041 2.207721953 0.500 11.3 11.29999927 2.388543915 0.625 12.725 12.72499895 2.686114145 0.750 14.15 14.38875508 3.106669977 0.875 15.575 15.57499968 3.673084269
1 17 17 5.313312737
We have calculated the weights of near optimum
portfolios under different values of coefficient of optimism by
applying equation 23. The weights are given in the following
Table 6.3.
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
159 | P a g e
Table 6.3: Weights of the near optimum portfolios of ten
securities under different α values
Securities
α values
0 0.125 0.250
0.375
0.500
0.625
0.750 0.875
1
1 0 0.1515975
0.303192
0.454787
0.606383
0.757979
0.9095745
0.7125
0
2 0 0 0 0 0 0 0 0.2875
1
3 0 0 0 0 0 0 0 0 0
4 0 0 0 0 0 0 0 0 0
5 0 0 0 0 0 0 0 0 0
6 0 0 0 0 0 0 0 0 0
7 0 0 0 0 0 0 0 0 0
8 0 0 0 0 0 0 0 0 0
9 0 0 0 0 0 0 0 0 0
10 1 0.8484043
0.696809
0.545213
0.393617
0.242021
0.0904255
0 0
The expected return and risk of the near optimum portfolios
under different co-efficient of optimism are shown in the
following Table 6.4:
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
160 | P a g e
Table 6.4: Expected return and risk of the near optimum
portfolios under different values of α
α values Minimum Return
Constraints
Expected Return
Risk
0 5.6 5.6 3.098386677
0.125 7.025 7.02502658 2.879043372
0.250 8.45 8.4500001 3.190095479
0.375 9.875 9.87499968 3.906859482
0.500 11.3 11.3000002 4.852771399
0.625 12.725 12.72499978 5.918970852
0.750 14.15 14.15000047 7.051102277
0.875 15.575 15.575 5.968700916
1 17 17 7.90569415
To observe the closeness between the Sharpe’s cut off
principle and the optimum portfolio under Sharpe’s Single Index
Model and the near optimum portfolio and the optimum portfolio
under Sharpe’s Single Index Model, we have considered City
Block Distance (CBD) to combine the two dimensions of
studying risk and return. The CBD between the optimum
portfolio under Sharpe’s Single Index Model and the Sharpe’s
cut off principle portfolio is given below in Table 6.5.
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
161 | P a g e
Table 6.5: CBD between the optimum portfolio under
Sharpe’s Single Index Model and Sharpe’s cut off principle
portfolio
α values Minimum Return Constraints
CBD values
0 5.6 9.900315476
0.125 7.025 9.664432223
0.250 8.45 8.324170016
0.375 9.875 6.757192409
0.500 11.3 5.151371587
0.625 12.725 3.428801677
0.750 14.15 1.344489715
0.875 15.575 0.408169178
1 17 3.473397966
The CBD between the optimum portfolios under Sharpe’s Single
Index Model and the near optimum portfolios are given in Table
6.6.
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
162 | P a g e
Table 6.6: CBD between the optimum portfolios under
Sharpe’s Single Index Model and the near optimum
portfolios
α values Minimum Return Constraints
CBD values
0 5.6 1.907376441
0.125 7.025 0.728587404
0.250 8.45 1.124350923
0.375 9.875 1.699138259
0.500 11.3 2.464228414
0.625 12.725 3.232857538
0.750 14.15 4.183186912
0.875 15.575 2.295616967
1 17 2.592381413
The CBD values between the optimum portfolio under
Sharpe’s Single Index Model and Sharpe’s cut off principle
based portfolio and between the optimum portfolio under
Sharpe’s Single Index Model and the near optimum portfolio are
graphically presented in the following Figure 6.1. Along the
horizontal axis we present the minimum return constraint and
along the vertical axis we present the CBD values.
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
163 | P a g e
Figure 6.1
From the above Figure 6.1 it is clear that to the left hand
side of the threshold point T1 (where the value of coefficient of
optimism α is 0.622), the distance between the near optimum
portfolio and the optimum portfolio under Sharpe’s Single Index
Model is less than that of Sharpe’s cut off principle portfolio. But
beyond the point T1 and up to point T2 (where the value of
coefficient of optimism is 0.962), Sharpe’s cut off principle
portfolio is closer to the optimum portfolio under Sharpe’s Single
Index Model. Again beyond point T2 the near optimum portfolio
Chapter 6: Comparative Analysis on Sharpe’s Cut off Principle and a Near Optimum Principle for Portfolio Selection under
Model Approach
164 | P a g e
is closer to the optimum portfolio under Sharpe’s Single Index
Model than that of the Sharpe’s cut off principle portfolio.
6.8 Conclusion
Based on the detailed empirical analysis, it may be
claimed that the closeness between the optimum portfolio under
Sharpe’s Single Index Model and Sharpe’s cut off principle
portfolio and the closeness between the optimum portfolio under
Sharpe’s Single Index Model and the near optimum portfolio
depends on the value of coefficient of optimism. When the value
of coefficient of optimism or α is in between 0 and 0.6, the near
optimum portfolio is closer to optimum portfolio under Sharpe’s
Single Index Model. But when the value of coefficient of
optimism (α) is in between point 0.6 and 0.95, the Sharpe’s Cut-
off principle portfolio is closer to the optimum portfolio under
Sharpe’s Single Index Model and when the value of coefficient
of optimism (α )is more than point 0.95 the near optimum
portfolio is again closer to the optimum portfolio under Sharpe’s
Single Index Model than that of Sharpe’s cut off principle
portfolio. Thus, we can say that up to moderate value and very
high value of coefficient of optimism, near optimum portfolio
shows better result. However, for moderate to high value of
coefficient of optimism, the Cut-off principle portfolio shows
closer result. This put forward the admissibility of the near
optimum portfolio.
V{tÑàxÜ J
XÇw exÅtÜ~á
Chapter 7: End Remarks
165 | P a g e
7.1 Importance of value system
Portfolio optimization is a balancing act between risk and
return, with each investor having unique requirements, as well
as unique financial outlook. However, the constant factor is the
delivery of an investment strategy which should not only be
specific to an investor's personal needs but also capable to work
well and provide financial security for the future. Within these
constraints the need arises to classify investors according to
their risk and return characteristics so that portfolio optimization
can be performed in order to suit individual investors. Every
investor is different, with different financial goals, different
tolerances to risk, different personal situations and different
desires.
The psychological dimension as provided in the realms of
behavioral finance shows that investors perceive gains and
losses asymmetrically, which boils down to the fact that
downside and upside movement by identical amounts does not
guarantee equal pleasure and pain. Further to it, investors tend
to modify their behavior according to their personal bias as well
as crowd pressure, thus making it difficult to predict the decision
making characteristics.
Chapter 7: End Remarks
166 | P a g e
Investors are not the same. Several dimensions including
personal and financial goals, current life situation and risk
tolerance tend to influence the differences among individual
investors. Investments that not only generate good returns but
also fulfill personal needs and provide financial security for the
future time horizon are usually preferred by the investors’
community at large. The investment decisions of the investors
are greatly influenced by the type of investor an individual is
(investment style). Value system plays a critical role in strategic
decision making. The concept has been extended to introduce
value system in order to classify investors and highlight their
performance in the portfolio optimization exercise. The following
four categories of investors are considered in this work
according to their value system.
Optimistic Investors, Risk Planner, Random Selector and
Pessimistic Investors
Optimistic investors have a propensity to take risk for
higher returns. They usually prefer risky investments and their
objective is to maximize returns in the long run. However, it is
observed that this type of investors seeks above average
returns by focusing investment in stocks and certain types of
mutual funds. Risk Planners seek to balance risk with assets
Chapter 7: End Remarks
167 | P a g e
that preserve the principal investment. However, they are not
totally risk averse as they are willing to take on some amount of
risk. Their investment strategy involves investing in portfolio of
securities which promises a constant income stream over the
investment horizon. The return objective of random selector is to
optimize return of the portfolio. As far as risk appetite is
concerned they tend to demonstrate a tolerable level as deemed
fit as per the need. The investment basket of random selectors
usually consists of equities or mutual funds and a mix of
moderate investments such as unit trusts and bonds. The
Pessimistic investors are categorized to be risk averse. The
investment principle of this type of investors is to preserve their
principal investment by avoiding risky investments. Thus,
government securities and bonds that promise a constant
income stream find prominent position in their investment
basket.
7.2 Closeness between optimum solution and
heuristic solution
As mathematical are used to generate optimum portfolio
are not easy to handle, heuristic solution is the other way to get
a near optimum portfolio. For those investors who find it difficult
Chapter 7: End Remarks
168 | P a g e
to mathematically or statistically arrive at sophisticated optimum
solution, they can heuristically obtain a portfolio similar to the
optimum portfolio. It means that one can reach to the near
optimum solution with the help of heuristic solution. In this
doctoral thesis, the heuristic solutions are already discussed
with empirical evidence. Detailed analysis revealed an ideal
point of change to be taken into consideration while examining
the behaviour of an investor, following heuristic approach in
investment planning. A mathematical formulation is presented to
design the optimum portfolio given a value for coefficient of
optimism.
After detailed discussion it is observed that heuristically
one can reach the optimum solution. In this doctoral thesis, the
concept of co-efficient of optimism is introduced in the decision
framework of the portfolio optimization and it is observed that
coefficient of optimism plays a crucial role in selecting portfolio
by considering human value system. Thus, investors have to be
aware of their coefficient of optimism.
When we introduced investor’s value system (according
to propensity to take risk) and coefficient of optimism in
choosing optimum portfolio heuristically, it is observed that the
optimistic investors and the risk planners are giving best
Chapter 7: End Remarks
169 | P a g e
decision. That means the risk planner’s decision is very close to
the optimum decision. For lower to moderate value of
coefficient of optimism risk planners’ decision and optimistic
investor’s decision is best as compared to the other heuristic
solutions. On the other hand, at the higher value of coefficient of
optimism all the heuristic solutions showed more or less same
result. For the lower to the moderate value of coefficient of
optimism random selectors showed worst result as compared to
the other heuristic solutions.
Next in this doctoral thesis, the Sharpe’s Single Index
model is reexamined and it is presented as a non-linear
mathematical model. A comparatively easy analytical framework
is achieved by modifying the Sharpe’s model which is called
near optimum portfolio. The modified decision rule is compared
with the Sharpe’s model Based on the detailed empirical
analysis, it may be claimed that the closeness between the
optimum portfolio under Sharpe’s Single Index Model and
Sharpe’s cut off principle portfolio and the closeness between
the optimum portfolio under Sharpe’s Single Index Model and
the near optimum portfolio depends on the value of coefficient of
optimism. Also by validating the models by with the selected
data set it is observed that from moderate value of coefficient of
Chapter 7: End Remarks
170 | P a g e
optimism to moderately high value of coefficient of optimism the
Sharpe’s Cut-off principle method gives the best result. For
lower value of coefficient of optimism to moderate value of
coefficient of optimism the near optimum portfolio gives the best
result. On the other hand, for a very high value of the coefficient
of optimism near optimum solution is also performing well. That
means an investor can reach near to the optimum portfolio
heuristically. This put forward the admissibility of the near
optimum portfolio.
7.3 Limitations of the Current work
The present work has aimed to reexamine the classical
model of portfolio optimization and has proposed new models of
portfolio optimization which when validated with the chosen data
set have demonstrated encouraging results. However the work
cannot be claimed to be robust in nature and readily usable for
commercial purpose due to the following limitations of the work.
1. Due to lack of time it was not possible to use
another data set in the work. So the work can be
defined as static in nature.
2. Further work can be done by considering the data of
other stock exchanges.
Chapter 7: End Remarks
171 | P a g e
3. Only risk parameter is considered in the work. Thus,
the scope of considering return parameter is there in
portfolio selection.
4. Only two portfolio theories have been considered in
the work. There is the scope of using other portfolio
theories.
7.4 Future Scope
Statistical models are not very easily understandable to
the general investor. When a general investor wishes to invest
money in any portfolio of securities they are more concerned
about the expected return and risk of the portfolio and not about
the various statistical models. Thus, the present study can be
further validated with other data set so as to make it robust and
thus ultimately lead to a simple heuristic tool to help investors so
that they can get a near optimum portfolio for investment.
Though heuristic method is not universally accepted but is
logically correct and thus the initiation in the present study can
be further made to be logically accurate with more insightful
research in the same dimension. The present study mainly
focuses on the weight of the securities in the portfolio; future
Chapter 7: End Remarks
172 | P a g e
studies can consider other dimensions in the portfolio
optimization framework. In real world, in an efficient market, it is
very difficult to predict assets’ weights accurately, thus, the
proposed model in the present study can be further explored
with more accurately predicted weights.
exyxÜxÇvxá
References
173 | P a g e
References
1) Arenas, M. et al (2004): Fuzzy compromise programming
for portfolio selection, MCDM, Whistler, B. C., Canada.
2) Beliakov, G. and Bagirov, A. (2006): Non –smooth
optimization methods for computation of the conditional
value at risk and portfolio optimization, Optimisation , Vol.
55, Issue 5 & 6, pp. 459 – 479.
3) Bilbao, A. et al (2006): An extension of Sharpe's single-
index model: portfolio selection with expert betas, Journal
of the Operational Research Society, Vol. 57, Number 12,
pp. 1442-1451(10)
4) Brandt, M. W. and Clara, P.S. (2006): Dynamic portfolio
selection by augmenting the asset space, The Journal of
Finance, VOL. LXI, NO. 5, pp. 2187-2217.
5) Britten-Jones, M. (1999): The sampling error in estimates
of mean-variance efficient portfolio weights, Journal of
Finance, Vol. 54, pp. 655-671.
6) Brown, S.J (1978): The portfolio choice problem:
comparison of certainty equivalence and optimal Bayes
References
174 | P a g e
portfolios. Commun.Statist.-Simula. Computa, B7, 321-
334.
7) Chang et. al (2000): Heuristics for cardinality constrained
portfolio optimization, Computers and Operation
Research,vol.27, issue 30, pp 1270-1302.
8) Chekhlov A., Uryasev, S. and Zabarankin, M. (2004):
Drawdown measure in portfolio optimization, Working
Paper, University of Florida.
9) Chitnis,A (2010), Performance Evaluation of Two Optimal
portfolios by Sharpe’s Ration , Global Journal of Financial
Management, Vol 2, pp. 35-46.
10) Copeland, Thomas E. and David Mayers ( 1982): The
value line enigma (1965-1978): A case study of
performance evaluation issues, Journal of Financial
Economics, 1982, 10 (3), 289-321.
11) Cornell, B. (1979): Asymmetric information and portfolio
performance measurement, Journal of Financial
Economics, 7(4), 381–390.
References
175 | P a g e
12) Cool T. (1999). Proper definitions for risk and
uncertainty,http://econpapers.repec.org/paper/wpawuwpg
t/9902002.htm.
13) Danielsson, Jorgensen J., Vries B., Yang C., Xiaoguang
(2008): Optimal portfolio allocation under the probabilistic
VaR constraint and incentives for financial innovation,
Annals of Finance.
14) Das, S.R. & Uppal, R.(2004): Systemic risk and
international portfolio choice, The Journal of Finance, Vol.
LIX, No. 6.
15) Detemple, J.B.Garcia, R. and Rindisbacher, M. (2003): A
Monte Carlo method for optimal portfolios, The Journal of
Finance, Vol. 58, Number 1, pp. 401-446(46).
16) Doganoglu, T., Hartz, C. and Mittnik, S. (2007): Portfolio
Optimization when risk factors are conditionally varying
and heavy tailed, Computational Economics.
17) Dutt, D.(2003): Constructing an optimal portfolio using
Sharpe’s Single Index Model, The Management
Accountant, VOL 38; PART 6, pages 452-455.
References
176 | P a g e
18) Elton, E.J. and Gruber, M.J. (1973): Modern portfolio
theory and investment analysis, John Wiley & Sons,
pp.128-137 and182-185.
19) Elton, E.J., Gruber, M.J. and Padberg, M.W (1978):
Optimal portfolios from simple ranking devices, Journal of
Portfolio Management, No.3, pp. 15-19.
20) Ferson, Wayne E. and Rudi W. Schadt (1996): Measuring
fund strategy and performancein changing economic
conditions, Journal of Finance, 51, 425-461.
21) Gibbons, M., Ross, S., and Shanken, J. (1989): A test of
the efficiency of a given portfolio, Econometrica, Vol. 57,
pp. 1121–1152.
22) Golosnoy, Vasyl, Okhrin Yarema (2007): Multivariate
shrinkage for optimal portfolio weights, The European
Journal of Finance,Vol. 13, Number 5, pp. 441-458(18).
23) Grinblatt, Mark and Sheridan Titman (1993): Performance
measurement without benchmarks: An examination of
mutual fund returns, Journal of Business, Vol.66, pp. 47-
68.
References
177 | P a g e
24) Heller,F. ( 1971): Managerial decision making; A study of
leadership style and power sharing among senior
managers, Tavistok Publication, London
25) Holton, Glyn A. (2004): Defining risk, Financial Analysts
Journal, Vol. 60 (6), pp. 19–25.
26) Huang, Xiaoxia (2008): Expected model for portfolio
selection with random fuzzy returns, International Journal
of General Systems.
27) Jauch, L. R. and Glueck, W. F. (1988): Business policy
and strategic management, McGraw-Hill International
Edition.
28) Jorion, P. (2000): Value at Risk: the new benchmark for
managing financial risk, McGraw-Hill.
29) Jones, M. B. (1999): The sampling error in estimates of
mean-variance efficient portfolio weights, The Journal of
Finance, Vol. LIV, No. 2.
30) Knight, F.H. (1921): Risk, uncertainty and profit, New
York Hart, Schaffner and Marx.
31) Ledoit, O. and Wolf, M.(2004): A well-conditioned
estimator for large-dimensional covariance matrices, The
References
178 | P a g e
Journal of Multivariate Analysis, vol.88, issue 2, pp365-
411.
32) Ledoit, O., and Wolf, M. (2004): Honey, I shrunk the
sample covariance matrix. Journal of Portfolio
Management, 30, 110–119.
33) Li Z.,Yang, H., Deng, X. (2007): Optimal dynamic portfolio
selection with Earnings- at- Risk, Journal of Optimisation
Theory and Applications.
34) Ma, L., Pohlman, L. (2008): Return forecasts and optimal
portfolio construction: a quantile regression approach,
SSRN working paper, SSRN-id880478.
35) Mansini, R. and Speranza, M.G. (1999): Heuristic
algorithms for the portfolio selection problem with
minimum transaction lots, European Journal of
Operational Research, 114, 219–233.
36) Markus E., Neumann, Throsten (2008): Time-varying
factor models for equity portfolio construction, The
European Journal of Finance.
37) Markowitz, H. M. (1952): Portfolio selection, The Journal
of Finance, Vol. 7, No. 1. Pp. 77-91.
References
179 | P a g e
38) Markowitz, H. M. (1959): Portfolio selection: efficient
diversification of investment, John Wiley. New York.
39) Merton, Robert C. (1969): Lifetime portfolio selection
under uncertainty: The continuous time case, Review of
Economics and Statistics, Vol. 51, pp. 247–257.
40) Merton, R. C.(1971): Optimum consumption and portfolio
rules in a continuous time model, Journal of Economic
Theory, Vol. 3, pp. 373–413.
41) Merton, Robert C. (1973): An intertemporal capital asset
pricing model, Econometrica, Vol.41, pp. 8678–87.
42) Merton, R.C. (1980): On estimating the expected return
on the market an exploratory investigation, Journal of
Financial Economics, Vol. 8, pp. 323-361.
43) Mishra, A. K. (2011): Construction of an optimal portfolio:
an application to Sharpe's Single Index model, Siddhant-
A Journal of Decision Making, Vol. 11, Issue 2, pp. 21-30.
44) Okhrin, Y. and Schmid, W. (2008): Estimation of optimal
portfolio weights, International Journal of Theoretical and
Applied Finance, Vol. 11, issue 03, pp. 249-276.
References
180 | P a g e
45) Rani, M., Bahl, S. (2012): Optimal portfolio selection with
or without the procedure of short sales, Asian Journal of
Research in Business Economics and Managemnet, Vol.
2, Issue 7.
46) Ross, S. A. (1976): The arbitrage pricing theory of capital
asset pricing, Journal of Economic Theory, Vol.13 (3), pp.
341–360.
47) Sahalia, Y. A. and Brandt, M.W. (2001): Variable
selection for portfolio choice, The Journal of Finance, Vol.
56, pp.1297-135.
48) Schaerf, A. (2002): Local search techniques for
constrained portfolio selection problems, Computational
Economics, Vol. 20(3), pp. 177–190.
49) Sharpe, W. F. (1963): A simplified model for portfolio
analysis, Management Science, pp. 277-293.
50) Sharpe, W. F. (1964): Capital asset prices: A theory of
market equilibrium under conditions of risk, Journal of
Finance, Vol. 19 (3), 425-442.
51) Sharpe, W. (1971): A linear programming approximation
for the general portfolio analysis problem, Journal of
References
181 | P a g e
Financial and Quantitative Analysis, December, pp. 1263-
1275.
52) Shaw, D.; Liu, S. ; Kopman, L. (2008): Lagrangian
relaxation procedure for cardinality-constrained portfolio
optimization, Optimization Methods and Software.
53) Stucchi, P. (1991): Some reflection about a simplified
algorithm of portfolio selection, 2nd AFIR Colloquium, 2:
323-337.
54) Taha, H.A.(1982): Operational research: an introduction,
McMillan, New York.
55) Tobin, J. (1958): Liquidity preference as behavior towards
risk, The Review of Economic Studies, Vol. 25, pp. 65-86.
56) Uppal, R., DeMiguel, V., Garlappi, L., & Nogales, F.
(2007): A generalized approach to portfolio optimization:
Improving performance by constraining portfolio norms
.SSRN Working Paper.
57) Ulucan, Aydın. (2007):An analysis of mean-variance
portfolio selection with varying holding periods, Applied
Economics.
References
182 | P a g e
WEBSITES
i. Wikipedia, Mathematical Model Approach (2012)
http://en.wikipedia.org/wiki/Mathematical_model
ii. Wikipedia, Non Linear Programming (2012)
http://en.wikipedia.org/wiki/Nonlinear_programming
accessed on June 2012.
iii. Wikipedia, Distance Measure (2012)
http://en.wikipedia.org/wiki/Distance accessed on June
2012.
iv. Wikipedia, Euclidean Distance (2012)
http://en.wikipedia.org/wiki/Euclidean_distance accessed
on June 2012.
v. Wikipedia, Mahalanobis Distance (2012)
http://en.wikipedia.org/wiki/Mahalanobis_distance
accessed on June 2012.
vi. Wikipedia, Manhattan Distance (2012)
http://en.wikipedia.org/wiki/Manhattan_distance accessed
on June 2012.
References
183 | P a g e
vii. Wikipedia, Chebyshev Distance (2012)
http://en.wikipedia.org/wiki/Chebyshev_distance
accessed on June 2012.
viii. Wikipedia, Minkowski Distance (2012)
http://en.wikipedia.org/wiki/Minkowski_distance accessed
on June 2012.
ix. Wikipedia, Linear Programming (2012)
http://en.wikipedia.org/wiki/Linear_programming
accessed on June 2012.
x. Wikipedia, Mathematical Optimization (2012)
http://en.wikipedia.org/wiki/Mathematical_optimization
accessed on June 2012.
xi. Wikipedia, Decision Theory (2012)
http://en.wikipedia.org/wiki/Decision_theory accessed on
June 2012.
xii. http://www.nseindia.com/products/dynaContent/equities/e
quities/histHighLowSearch.htm accessed on June 2010.