PORTFOLIO THEORY

The Benefits of Diversification

 Outline
• Diversification and Portfolio Risk

• Portfolio Return and Risk

• Measurement of Co movements in Security Returns

• Calculation of Portfolio Risk

• Efficient Frontier

• Optimal Portfolio

• Riskless Lending and Borrowing

• The Single Index Model

 Diversification and Portfolio Risk
Before we look at the formula for portfolio risk, let us understand somewhat
intuitively how diversification influences risk. Suppose you have Rs.100,000 to
invest and you want to invest it equally in two stocks, A and B. The return on these
stocks depends on the state of the economy. Your assessment suggests that the
probability distributions of the returns on stocks A and B are as shown in Exhibit
7.1. For the sake of simplicity, all the five states of the economy are assumed to be
equiprobable. The last column of Exhibit 7.2 shows the return on a portfolio
consisting of stocks A and B in equal proportions. Graphically, the returns are
shown in Exhibit 7.2.
 
 
Probability Distribution of Returns

 State of the Probability Return on Return on Return on

Economy Stock A Stock B Portfolio
 
1 0.20 15% -5% 5%
2 0.20 -5% 15 5%
3 0.20 5 25 15%
4 0.20 35 5 20%
5 0.20 25 35 30%
Returns on Individual Stocks and the
Portfolio
Expected Return and Standard Deviation
Expected Return

Stock A : 0.2(15%) + 0.2(-5%) + 0.2(5%) +0.2(35%) + 0.2(25%) = 15%

Stock B : 0.2(-5%) + 0.2(15%) + 0.2(25%) + 0.2(5%) + 0.2(35%) = 15%
Portfolio of
A and B : 0.2(5%) + 0.2(5%) + 0.2(15%) + 0.2(20%) + 0.2(30%) = 15%

Standard Deviation

Stock A : σ2A = 0.2(15-15)2 + 0.2(-5-15)2 + 0.2(5-15)2 + 0.2(35-15)2 + 0.20

(25-15)2
= 200
σA = (200)1/2 = 14.14%
Stock B : σ2B = 0.2(-5-15)2 + 0.2(15-15)2 + 0.2(25-15)2 + 0.2(5-15)2 + 0.2
(35-15)2
= 200
σB = (200)1/2 = 14.14%
Portfolio : σ2(A+B) = 0.2(5-15)2 + 0.2(5-15)2 + 0.2(15-15)2 + 0.2(20-15)2
+ 0.2(30-15)2
= 90
σA+B = (90)1/2 = 9.49%
 Portfolio Expected Return
n
E(RP) =  wi E(Ri)
i=1
where E(RP) = expected portfolio return
wi = weight assigned to security i
E(Ri) = expected return on security i
n = number of securities in the portfolio
Example A portfolio consists of four securities with expected returns
of 12%, 15%, 18%, and 20% respectively. The proportions of
portfolio value invested in these securities are 0.2, 0.3, 0.3, and 0.20
respectively.
The expected return on the portfolio is:
E(RP) = 0.2(12%) + 0.3(15%) + 0.3(18%) + 0.2(20%)
= 16.3%
Relationship Between Diversification and
Risk
Risk

Unique

Risk

Market Risk

1 5 10 20 No. of Securities
 Market Risk Versus Unique Risk

• Basic insight of modern portfolio theory:

Total risk = Unique risk + Market risk

• The unique risk of a security represents that portion of it total

risk which stems from firm-specific factors.

• The Market risk of a stock represents that portion of its risk

which is attribute to economy wide factors
 Equations for Portfolio Risk
In symbols

E(Rp) =  wi E(Ri) i=1

 
But
p2   wi 2i 2

Thanks to the inequality shown in Eq. (7.1), investors can achieve

the benefit of risk reduction through diversification. Before we
discuss how this can be accomplished let us first understand how
comovements in security returns are measured

 
 Portfolio Risk
The risk of a portfolio is measured by the variance (or standard
deviation) of its return. Although the expected return on a portfolio
is the weighted average of the expected returns on the individual
securities in the portfolio, portfolio risk is not the weighted average
of the risks of the individual securities in the portfolio (except when
the returns from the securities are uncorrelated).
 Measurement Of Comovements
In Security Returns
• To develop the equation for calculating portfolio risk we need
information on weighted individual security risks and
weighted comovements between the returns of securities
included in the portfolio.
• Comovements between the returns of securities are measured
by covariance (an absolute measure) and coefficient of
correlation (a relative measure).
 Covariance

COV (Ri , Rj) = p1 [Ri1 – E(Ri)] [ Rj1 – E(Rj)]

+ p2 [Ri2 – E(Rj)] [Rj2 – E(Rj)]

+ ••
••

+ pn [Rin – E(Ri)] [Rjn – E(Rj)]

 Illustration
The returns on assets 1 and 2 under five possible states of nature are given below

State of nature Probability Return on asset 1 Return on asset 2

1 0.10 -10% 5%
2 0.30 15 12
3 0.30 18 19
4 0.20 22 15
5 0.10 27 12

The expected return on asset 1 is :

E(R1) = 0.10 (-10%) + 0.30 (15%) + 0.30 (18%) + 0.20 (22%) + 0.10 (27%) = 16%

The expected return on asset 2 is :

E(R2) = 0.10 (5%) + 0.30 (12%) + 0.30 (19%) + 0.20 (15%) + 0.10 (12%) = 14%

The covariance between the returns on assets 1 and 2 is calculated below :

State of Probability Return on Deviation of Return on Deviation of Product of the
nature asset 1 the return on asset 2 the return on deviations
asset 1 from its asset 2 from times
mean its mean probability
(1) (2) (3) (4) (5) (6) (2) x (4) x (6)

1 0.10 -10% -26% 5% -9% 23.4

2 0.30 15% -1% 12% -2% 0.6
3 0.30 18% 2% 19% 5% 3.0
4 0.20 22% 6% 15% 1% 1.2
5 0.10 27% 11% 12% -2% -2.2
Sum = 26.0

Thus the covariance between the returns on the two assets is 26.0.
 Coefficient Of Correlation
Cov (Ri , Rj)
Cor (Ri , Rj) or ij =
ij
ij
=
 i j
ij = ij . i . j
where ij = correlation coefficient between the returns on
securities i and j
ij = covariance between the returns on securities
i and j
i , j = standard deviation of the returns on securities
i and j
Graphical Portrayal of Various Types of
Correlation Relationships
 Portfolio Risk : 2 – Security Case
p = [w12 12 + w22 22 + 2w1w2 12 1 2]½

Example : w1 = 0.6 , w2 = 0.4,

1 = 10%, 2 = 16%

12 = 0.5

p = [0.62 x 102 + 0.42 x 162 +2 x 0.6 x 0.4 x 0.5 x 10 x 16]½

= 10.7%
 Portfolio Risk : n – Security Case
p = [   wi wj ij i j ] ½
Example : w1 = 0.5 , w2 = 0.3, and w3 = 0.2
1 = 10%, 2 = 15%, 3 = 20%
12 = 0.3, 13 = 0.5, 23 = 0.6
p = [w12 12 + w22 22 + w32 32 + 2 w1 w2 12 1 2
+ 2w2 w3 13 1 3 + 2w2 w3 232 3] ½
= [0.52 x 102 + 0.32 x 152 + 0.22 x 202
+ 2 x 0.5 x 0.3 x 0.3 x 10 x 15
+ 2 x 0.5 x 0.2 x 05 x 10 x 20
+ 2 x 0.3 x 0.2 x 0.6 x 15 x 20] ½
= 10.79%
 Risk Of An N - Asset Portfolio

2p =   wi wj ij i j

1n x n MATRIX
2 3 … n

1 w12σ12 w1w2ρ12σ1σ2 w1w3ρ13σ1σ3 … w1wnρ1nσ1σn

2 w2w1ρ21σ2σ1 w22σ22 w2w3ρ23σ2σ3 … w2wnρ2nσ2σn

3 w3w1ρ31σ3σ1 w3w2ρ32σ3σ2 w32σ32 …

: : :

n wnw1ρn1σnσ1 wn2σn2
 Dominance Of Covariance

As the number of securities included in a portfolio increases, the

importance of the risk of each individual security decreases whereas

the significance of the covariance relationship increases.

Quantitative Expression for Dominance of Covariance
1 2 3 … n
1 w12σ12 w1w2ρ12σ1σ2 W1w3ρ13σ1σ3 … w1wnρ1nσ1σn

w2w1ρ21σ2σ1 w22σ22 W2w3ρ23σ2σ3 … w2wnρ2nσ2σn

2 3 …
w3w1ρ31σ3σ1 w3w2ρ32σ3σ2 W32σ32

3 : :
: wn2σn2
wnw1ρn1σnσ1
n
Var(Rp) = Σ w12 Var(Ri) + Σ Σ wi wj Cov (Ri, Rj) i=1 i=1 j=1
i¹ j
 
If a naïve diversification strategy is followed wi = 1/n. Under such a strategy
  n n n
Var (Rp) =1/n Σ 1/n Var (Ri) + Σ Σ 1/n2 Cov (Ri, Rj)
i=1 i=1 j=1
j¹i
 The average variance term and the average covariance term may be expressed as follows:
  n
Var = 1/n Σ Var (Ri)
i=1
  1 n n
Cov = Σ Σ Cov (Ri, Rj)
n(n-1) i=1 j=1
i¹j
Hence
1 n-1
Var (Rp) = Var + Cov n n
 
As n increases, the first term tends to become zero and the second term looms large. Put differently, the
importance of the variance term diminishes whereas the importance of the covariance term increases.
Efficient Frontier
For A Two Security-Case
Security A Security B
Expected return 12% 20%
Standard deviation 20% 40%
Coefficient of correlation -0.2

Portfolio Proportion of A Proportion of B Expected return Standard deviation

wA wB E (Rp) p
1 (A) 1.00 0.00 12.00% 20.00%

2 0.90 0.10 12.80% 17.64%

3 0.759 0.241 13.93% 16.27%

4 0.50 0.50 16.00% 20.49%

5 0.25 0.75 18.00% 29.41%

6 (B) 0.00 1.00 20.00% 40.00%

 Portfolio Options And
The Efficient Frontier
Expected
return , E(Rp)

20% 6 (B)

3•
2•
12% 1 (A)

Risk, p
20% 40%
Feasible Frontier Under Various Degrees Of
Coefficient of Correlation
Expected
return , E (Rp)

20% B (WB = 1)

0
=
  = 1 .0
•
=–
1 .0
12% A (WA = 1)

Standard deviation, p
 Minimum Variance Portfolio
Most investors (and portfolio managers) invest in two broad
categories of financial assets viz., bonds and stocks. So, an
important practical issue is: what is the proportion of bonds (and, by
derivation, stocks) that minimises portfolio variance? To answer this
question, let us look at the risk of a portfolio consisting of two assets,
viz., bonds and stocks:
 
Var (Rp) = w2bσ2b + w2sσ2s + 2wbwS ρbsσBσS
 where Var (Rp) is the variance of the portfolio consisting of bonds
and stocks, wb is the proportion invested in bonds, wS is the
proportion invested in stocks (wS =1-wb), σB is the standard deviation
of returns from bonds, σS is the standard deviation of returns from
stocks, and ρbs is the coefficient of correlation between the returns
from bonds and stocks.
 Minimum Variance Portfolio -2
The value of wb that minimises portfolio variance is:

σ2s - σbσS ρbS

wB (min) =
σ2b+ σ2s - 2σbσS ρbS

To illustrate the above formula, let us consider the following data: E(rB) = 8%,
E(rs) = 15%, σB = 0%, σS = 20%.
Given the above data the expected return and standard deviation of a portfolio
consisting of bonds and stocks is:

E(rp) = wb . 8 + ws . 15
 
σp = [ w2b . 100 + w2s . 400 + 2wb.ws.pbs . 200 ]1/2
 
where E(rp) is the expected portfolio return, σp is the portfolio standard deviation,
wb and ws are the proportions invested in bonds and stocks, ρBS is the coefficient of
correlation between the returns on bonds and stocks.
 Minimum Variance Portfolio -3

The minimum portfolio variance for various correlations is shown

below
Correlation

ρ = - 1.0 ρ=0 ρ = 0.5

Minimum Variance Portfolio

wb (min) 0.6667 0.8000 1.000

E (rp) 10.33% 9.400 8.000
σp 0 12.00% 10.00%
Efficient Frontier For The
n-Security Case
Expected
return , E (Rp)

•X
•
F D
• B Z• •M

•N
•O
A
•

Standard deviation, p
 Approaches to Determining the Efficient
Frontier

• Graphical Analysis

• Calculus Analysis

• Quadratic Programming Analysis

 Quadratic Programming Analysis

Technically, the quadratic programming approach manipulates the

portfolio weights to determine efficient portfolios. The procedure
followed is as follows. A desired expected return, say 9 percent, is
specified. Then all portfolios (combinations of securities) that
produce 9 percent expected returns are considered and the portfolio
that has the smallest variance (standard deviation) of return is
chosen as the efficient portfolio. This is continued for other levels of
portfolio return, 10 percent, 11 percent, 12 percent, and so on, until
all the possible expected returns are considered. Alternatively, the
problem can be solved by specifying various levels of portfolio
variance (standard deviation) and choosing the portfolios that offer
the highest expected return for various levels of portfolio variance
(standard deviation).
 Why Is the Feasible Region Broken-egg
Shaped-1
To see why the feasible region has a broken-egg shape (or umbrella shape), let us look at
two securities shown in Exhibit 7.10.

Exhibit 7.10  Two Securities Portfolio

Expected
return, E(r)

Standard deviation, σ
 Why Is the Feasible Region Broken-egg
Shaped-2

The expected return of a portfolio comprising of A and B (wa + wb = 1) is:

E(RP) = wa E(Ra) + wb E(Rb)

The standard deviation of the portfolio return is:

σ P= [wa2 σ a2 + wb2 σ b2 + 2 wa wbρab σ a σ b ]½

Note that while the expected portfolio return is not affected by Pab, the coefficient
of correlation between the returns on A and B, the standard deviation of
portfolio return is affected by Pab.
 Why Is the Feasible Region Broken-egg
Shaped-3
Now consider two cases:
Case 1:  The returns of securities A and B are perfectly positively correlated.
Case 2:  The returns of securities A and B are less than perfectly positively correlated.

In case 1, ρab, = 1.
So σ p = [wa2 σ a2 + wb2 σ b2 + 2 wa wb σ a σ b ]½
= [waσ a + wb σ b]

Graphically, this means that the portfolio risk-return profile plots as the straight line joining A and B in
Exhibit 7.10.
In case 2, ρab, < 1.
So
σ p = [wa2 σ a2 + wb2 σ b2 + 2 wawb ρab σ a σ b ]½

Since ρab < 1, σ p will be less than [wa σ a + wb σb]

profile plots as the broken curved line joining A and B in Exhibit 7.10.
Since ρab is typically less than 1, most of the portfolio risk-return profiles are like the curved line. This
implies that the feasible region would have an umbrella like shape.
Risk-Return Indifference Curves

IQ
IP
Expected
return, E(Rp)

Standard

deviation σ p,
 Utility Indifferences Curves

Expected
return, E(Rp)
Ip4
Ip3
Ip2 Ip1

A Standard

deviation σ p,
 Optimal Portfolio
Expected
return , E (Rp) IP3 I I
IQ2 P2 P1
IQ3 IQ1
•X
• *
P
•
Q* • M

•
N

•
O

Standard deviation, p
 Riskless Lending And
Expected
Borrowing Opportunity
return , E (Rp)
G
II
V
E •
• • •X
S
• B I

D •M
• •F Y
C• •
u •N
•
Rf • •O
•
A

Standard deviation, p

Thus, with the opportunity of lending and borrowing, the efficient frontier changes. It is no
longer AFX. Rather, it becomes Rf SG as it domniates AFX.
 Separation Theorem
• Since Rf SG dominates AFX, every investor would do well to
choose some combination of Rf and S. A conservative investor
may choose a point like U, whereas an aggressive investor
may choose a point like V.

• Thus, the task of portfolio selection can be separated into two

steps:
1. Identification of S, the optimal portfolio of risky
securities.
2. Choice of a combination of Rf and S, depending on one’s
risk attitude.
This is the import of the celebrated separation theorem
 Single Index Model
Information - Intensity of the Markowitz model n Securities
n Variance terms & n(n -1) /2
Covariance terms

Sharpe’s Model
Rit = ai + bi RMt + eit
E(Ri) = ai + bi E(RM)
var (Ri) = bi2 [var (RM)] + var (ei)
cov (Ri ,Rj) = bi bj var (RM)

Markowitz Model Single Index Model

n (n + 3)/2 3n + 2
E (Ri) & var (Ri) for each bi , bj var (ei) for
security n( n - 1)/2 each security &
covariance terms E (RM) & var (RM)
 Summing up
• Portfolio theory, originally proposed by Markowitz, is the
first formal attempt to quantify the risk of a portfolio and
develop a methodology for determining the optimal portfolio.

• The expected return on a portfolio of n securities is :

E(Rp) =  wi E(Ri)

• The variance and standard deviation of the return on an

n-security portfolio are:
p2 =   wi wj ij i j
p =   wi wj ij i j ½

• A portfolio is efficient if (and only if) there is an alternative

with (i) the same E(Rp) and a lower p or (ii) the same p and
a higher E(Rp), or (iii) a higher E(Rp) and a lower p
• Given the efficient frontier and the risk-return indifference
curves, the optimal portfolio is found at the tangency between
the efficient frontier and a utility indifference curve.

• If we introduce the opportunity for lending and borrowing at

the risk-free rate, the efficient frontier changes dramatically.
It is simply the straight line from the risk-free rate which is
tangential to the broken-egg shaped feasible region
representing all possible combinations of risky assets.

• The Markowitz model is highly information-intensive.

• The single index model, proposed by Sharpe, is a very helpful
simplification of the Markowitz model.