Econ 424/Amath 540

Introduction to Portfolio Theory

Eric Zivot
July 31, 2012
Introduction to Portfolio Theory
Investment in Two Risky Assets
1

= simple return on asset A

1
1
= simple return on asset B
W
0
= initial wealth
Assumptions
1

and 1
1
are described by the CER model
1
i
iid .(j
i
, o
2
i
), i = , 1
cov(1

, 1
1
) = o
1
, cor(1

, 1
1
) = j
1
Investors like high 1[1
i
] = j
i
Investors dislike high var(1
i
) = o
2
i
Investment horizon is one period (e.g., one month or one year)
Note: Traditionally in portfolio theory, returns are simple and not continuously
compounded
Portfolios
a

= share of wealth in asset A =

$ in A
W
0
a
1
= share of wealth in asset B =
$ in B
W
0
Long position
a

, a
1
0
Short position
a

< 0 or a
1
< 0
Assumption: Allocate all wealth between assets A and B
a

+ a
1
= 1
Portfolio Return
1
j
= a

+ a
1
1
1
Portfolio Distribution
j
j
= 1[1
j
] = a

+ a
1
j
1
o
2
j
= var(1
j
) = a
2

o
2

+ a
2
1
o
2
1
+ 2a

a
1
o
1
= a
2

o
2

+ a
2
1
o
2
1
+ 2a

a
1
j
1
o

o
1
1
j
iid .(j
j
, o
2
j
)
End of Period Wealth
W
1
= W
0
(1 + 1
j
) = W
0
(1 + a

+ a
1
1
1
)
W
1
.(W
0
(1 + j
j
), o
2
j
W
2
0
)
Example Data
j

= 0.175, j
1
= 0.055
o
2

= 0.067, o
2
1
= 0.013
o

= 0.258, o
1
= 0.115
o
1
= 0.004875,
j
1
=
o
1
o

o
1
= 0.164
Note: Asset A has higher expected return and risk than asset B
Example: Long only two asset portfolio
Consider an equally weighted portfolio with a

= a
1
= 0.5. The expected
return, variance and volatility are
j
j
= (0.5) (0.175) + (0.5) (0.055) = 0.115
o
2
j
= (0.5)
2
(0.067) + (0.5)
2
(0.013)
+ 2 (0.5)(0.5)(0.004875) = 0.01751
o
j
=

0.01751 = 0.1323
This portfolio has expected return half-way between the expected returns on as-
sets A and B, but the portfolio standard deviation is less than half-way between
the asset standard deviations. This reects risk reduction via diversication.
Example: Long-Short two asset portfolio
Next, consider a long-short portfolio with a

= 1.5 and a
1
= 0.5. In this
portfolio, asset B is sold short and the proceeds of the short sale are used to
leverage the investment in asset A. The portfolio characteristics are
j
j
= (1.5) (0.175) + (0.5) (0.055) = 0.235
o
2
j
= (1.5)
2
(0.067) + (0.5)
2
(0.013)
+ 2 (1.5)(0.5)(0.004875) = 0.1604
o
j
=

0.01751 = 0.4005
This portfolio has both a higher expected return and standard deviation than
asset A
Portfolio Value-at-Risk
Assume an initial investment of $W
0
in the portfolio of assets A and B.
Given that the simple return 1
j
.(j
j
, o
2
j
), For c (0, 1), the c
100% portfolio value-at-risk is
VaR
j,c
= q
1
j,c
W
0
= (j
j
+ o
j
q
:
c
) W
0
where q
1
j,c
is the c quantile of the distribution of 1
j
and q
:
c
= c quantile
of Z .(0, 1).
Relationship between Portfolio VaR and Individual Asset VaR
Result: Portfolio VaR is not a weighted average of asset VaR
VaR
j,c
6= a

VaR
,c
+ a
1
VaR
1,c
unless j
1
= 1.
Asset VaRs for A and B are
VaR
,c
= q
1

0.05
W
0
= (j

+ o

q
:
c
)W
0
VaR
1,c
= q
1
1
0.05
W
0
= (j
1
+ o
1
q
:
c
)W
0
Portfolio VaR is
VaR
j,c
= (j
j
+ o
j
q
:
c
) W
0
=

(a

+ a
1
j
1
) +

a
2

o
2

+ a
2
1
o
2
1
+ 2a

a
1
o
1

12
q
:
c

W
0
Portfolio weighted asset VaR is
a

VaR
,c
+ a
1
VaR
1,c
= a

(j

+ o

q
:
c
)W
0
+ a
1
(j
1
+ o
1
q
:
c
)W
0
= [(a

+ a
1
j
1
) + (a

+ a
1
o
1
) q
:
c
] W
0
6= (j
j
+ o
j
q
:
c
) W
0
= VaR
j,c
provided j
1
6= 1.
If j
1
= 1 then o
1
= j
1
o

o
1
= o

o
1
and
o
2
j
= a
2

o
2

+ a
2
1
o
2
1
+ 2a

a
1
o

o
1
= (a

+ a
1
o
1
)
2
o
j
= a

+ a
1
o
1
and so
a

VaR
,c
+ a
1
VaR
1,c
= VaR
j,c
Example: Portfolio VaR and Individual Asset VaR
Consider an initial investment of W
0
=$100,000. The 5% VaRs on assets A
and B are
VaR
,0.05
= q
1

0.05
W
0
= (0.175 + 0.258(1.645)) 100, 000 = 24, 937,
VaR
1,0.05
= q
1
1
0.05
W
0
= (0.055 + 0.115(1.645)) 100, 000 = 13, 416.
The 5% VaR on the equal weighted portfolio with a

= a
1
= 0.5 is
VaR
j,0.05
= q
1
j
0.05
W
0
= (0.115 + 0.1323(1.645)) 100, 000 = 10, 268,
and the weighted average of the individual asset VaRs is
a

VaR
,0.05
+a
1
VaR
1,0.05
= 0.5(24, 937)+0.5(13, 416) = 19, 177.
Portfolio Frontier
Vary investment shares a

and a
1
and compute resulting values of j
j
and
o
2
j
. Plot j
j
against o
j
as functions of a

and a
1
Shape of portfolio frontier depends on correlation between assets A and B
If j
1
= 1 then there exists portfolio shares a

and a
1
such that
o
2
j
= 0
If j
1
= 1 then there is no benet from diversication
Diversication is benecial even 0 < j
1
< 1
Ecient Portfolios
Denition: Portfolios with the highest expected return for a given level of risk,
as measured by portfolio standard deviation, are ecient portfolios
If investors like portfolios with high expected returns and dislike portfolios
with high return standard deviations then they will want to hold ecient
portfolios
Which ecient portfolio an investor will hold depends on their risk prefer-
ences
Very risk averse investors dislike volatility and will hold portfolios near
the global minimum variance portfolio. They sacrice expected return
for the safety of low volatility
Risk tolerant investors dont mind volatility and will hold portfolios that
have high expected returns. They gain expected return by taking on
more volatility.
Globabl Minimum Variance Portfolio
The portfolio with the smallest possible variance is called the global mini-
mum variance portfolio.
This portfolio is chosen by the most risk averse individuals
To nd this portfolio, one has to solve the following constrained minimiza-
tion problem
min
a

,a
1
o
2
j
= a
2

o
2

+ a
2
1
o
2
1
+ 2a

a
1
o
1
c.t. a

+ a
1
= 1
Review of Optimization Techniques: Constrained Optimization
Example: Finding the minimum of a bivariate function subject to a linear con-
straint
j = )(a, :) = a
2
+ :
2
min
a,:
j = )(a, :)
c.t. a + : = 1
Solution methods:
Substitution
Lagrange multipliers
Method of Substitution
Substitute : = a 1 in )(a, :) and solve univariate minimization
j = )(a, a 1) = a
2
+ (1 a)
2
min
a
)(a, a 1)
First order conditions
0 =
o
oa
(a
2
+ (1 a)) = 2a + 2(1 a)(1)
= 4a 2
a = 0.5
Solving for :
: = 1 0.5 = 0.5
Method of Lagrange Multipliers
Idea: Augment function to be minimized with extra terms to impose constraints
1. Put constraints in homogeneous form
a + : = 1 a + : 1 = 0
2. Form Lagrangian function
1(a, :, A) = a
2
+ :
2
+ A(a + : 1)
A = Lagrange multiplier
3. Minimize Lagrangian function
min
a,:,A
1(a, :, A)
First order conditions
0 =
01(a, :, A)
0a
= 2 a + A
0 =
01(a, :, A)
0:
= 2 : + A
0 =
01(a, :, A)
0A
= a + : 1
We have three linear equations in three unknowns. Solving gives
2a = 2: = A a = :
2: 1 = 0 : = 0.5, a = 0.5
Example: Finding the Global Minimum Variance Portfolio
Two methods for solution
Analytic solution using Calculus
Numerical solution
use the Solver in Excel
use R function solve.QP() in package quadprog for quadratic opti-
mization problems with equality and inequality constraints
Calculus Solution
Minimization problem
min
a

,a
1
o
2
j
= a
2

o
2

+ a
2
1
o
2
1
+ 2a

a
1
o
1
c.t. a

+ a
1
= 1
Use substitution method with
a
1
= 1 a

to give the univariate minimization

min
a

o
2
j
= a
2

o
2

+ (1 a

)
2
o
2
1
+ 2a

(1 a

)o
1
First order conditions
0 =
o
oa

o
2
j
=
o
oa

a
2

o
2

+ (1 a

)
2
o
2
1
+ 2a

(1 a

)o
1

= 2a

o
2

2(1 a

)o
2
1
+ 2o
1
(1 2a

)
a
min

=
o
2
1
o
1
o
2

+ o
2
1
2o
1
, a
min
1
= 1 a
min

Excel Solver Solution

The Solver is an Excel add-in, that can be used to numerically solve general
linear and nonlinear optimization problems subject to equality or inequality
constraints
The solver is made by FrontLine Systems and is provided with Excel
The solver add-in may not be installed in a default installation of Excel
Tools/Add-Ins and check the Solver Add-In box
If Solver Add-In box is not available, the Solver Add-In must be installed
from original Excel installation CD
Portfolios with a Risk Free Asset
Risk Free Asset
Asset with xed and known rate of return over investment horizon
Usually use U.S. government T-Bill rate (horizons < 1 year) or T-Note
rate (horizon 1 yr)
T-Bill or T-Note rate is only nominally risk free
Properties of Risk-Free Asset
1
)
= return on risk-free asset
1[1
)
] = v
)
= constant
var(1
)
) = 0
cov(1
)
, 1
i
) = 0, 1
i
= return on any asset
Portfolios of Risky Asset and Risk Free Asset
a
)
= share of wealth in T-Bills
a
1
= share of wealth in asset B
a
)
+ a
1
= 1
a
)
= 1 a
1
Portfolio return
1
j
= a
)
v
)
+ a
1
1
1
= (1 a
1
)v
)
+ a
1
1
1
= v
)
+ a
1
(1
1
v
)
)
Portfolio excess return
1
j
v
)
= a
1
(1
1
v
)
)
Portfolio Distribution
j
j
= 1[1
j
] = v
)
+ a
1
(j
1
v
)
)
o
2
j
= var(1
j
) = a
2
1
o
2
1
o
j
= a
1
o
1
1
j
.(j
j
, o
2
j
)
Risk Premium
j
1
v
)
= excess expected return on asset B
= expected return on risky asset over return on safe asset
For the portfolio of T-Bills and asset B
j
j
v
)
= a
1
(j
1
v
)
)
= expected portfolio return over T-Bill
The risk premia is an increasing function of the amount invested in asset B.
Leveraged Investment
a
)
< 0, a
1
1
Borrow at T-Bill rate to buy more of asset B
Result: Leverage increases portfolio expected return and risk
j
j
= v
)
+ a
1
(j
1
v
)
)
o
j
= a
1
o
1
a
1
j
j
& o
j

Determining Portfolio Frontier

Goal: Plot j
j
vs. o
j
o
j
= a
1
o
1
a
1
=
o
j
o
1
j
j
= v
)
+ a
1
(j
1
v
)
)
= v
)
+
o
j
o
1
(j
1
v
)
)
= v
)
+

j
1
v
)
o
1
!
o
j
where

j
1
v
)
o
1
!
= SR
1
= Asset B Sharpe Ratio
= excess expected return per unit risk
Remarks
The Sharpe Ratio (SR) is commonly used to rank assets.
Assets with high Sharpe Ratios are preferred to assets with low Sharpe
Ratios
Ecient Portfolios with 2 Risky Assets and a Risk Free Asset
Investment in 2 Risky Assets and T-Bill
1

= simple return on asset A

1
1
= simple return on asset B
1
)
= v
)
= return on T-Bill
Assumptions
1

and 1
1
are described by the CER model
1
i
iio .(j
i
, o
2
i
), i = , 1
cov(1

, 1
1
) = o
1
, corr(1

, 1
1
) = j
1
Results:
The best portfolio of two risky assets and T-Bills is the one with the highest
Sharpe Ratio
Graphically, this portfolio occurs at the tangency point of a line drawn
from 1
)
to the risky asset only frontier.
The maximum Sharpe Ratio portfolio is called the tangency portfolio
Mutual Fund Separation Theorem
Ecient portfolios are combinations of two portfolios (mutual funds)
T-Bill portfolio
Tangency portfolio - portfolio of assets A and B that has the maximum
Shape ratio
Implication: All investors hold assets A and B according to their proportions in
the tangency portfolio regardless of their risk preferences.
Finding the tangency portfolio
max
a

, a
1
SR
j
=
j
j
v
)
o
j
subject to
j
j
= a

+ a
1
j
1
o
2
j
= a
2

o
2

+ a
2
1
o
2
1
+ 2a

a
1
o
1
1 = a

+ a
1
Solution can be found analytically or numerically (e.g., using solver in Excel)
Using the substitution method it can be shown that
a
tan

=
(j

v
)
)o
2
1
(j
1
v
)
)o
1
(j

v
)
)o
2
1
+ (j
1
v
)
)o
2

(j

v
)
+ j
1
v
)
)o
1
a
tan
1
= 1 a
tan

Portfolio characteristics
j
tan
j
= a
tan

+ a
tan
1
j
1

o
tan
j

2
=

a
tan

2
o
2

+

a
tan
1

2
o
2
1
+ 2a
tan

a
tan
1
o
1
Ecient Portfolios: tangency portfolio plus T-Bills
a
tan
= share of wealth in tangency portfolio
a
)
= share of wealth in T-bills
a
tan
+ a
)
= 1
j
c
j
= v
)
+ a
tan
(j
tan
j
v
)
)
o
c
j
= a
tan
o
tan
j
Result: The weights a
tan
and a
)
are determined by an investors risk prefer-
ences
Risk averse investors hold mostly T-Bills
Risk tolerant investors hold mostly tangency portfolio
Example
For the two asset example, the tangency portfolio is
a
tan

= .46, a
tan
1
= 0.54
j
tan
j
= (.46)(.175) + (.54)(.055) = 0.11

o
tan
j

2
= (.46)
2
(.067) + (.54)
2
(.013)
+ 2(.46)(.54)(.005)
= 0.015
o
tan
j
=

.015 = 0.124
Ecient portfolios have the following characteristics
j
c
j
= v
)
+ a
tan
(j
tan
j
v
)
)
= 0.03 + a
tan
(0.11 0.03)
o
c
j
= a
tan
o
tan
j
= a
tan
(0.124)
Problem: Find the ecient portfolio that has the same risk (SD) as asset B?
That is, determine a
tan
and a
)
such that
o
c
j
= o
1
= 0.114 = target risk
Note: The ecient portfolio will have a higher expected return than asset B
Solution:
.114 = o
c
j
= a
tan
o
tan
j
= a
tan
(.124)
a
tan
=
.114
.124
= .92
a
)
= 1 a
tan
= .08
Ecient portfolio with same risk as asset B has
(.92)(.46) = .42 in asset A
(.92)(.54) = .50 in asset B
.08 in T-Bills
If v
)
= 0.03, then expected Return on ecient portfolio is
j
c
j
= .03 + (.92)(.11 0.03) = .104
Problem: Assume that v
)
= 0.03. Find the ecient portfolio that has the
same expected return as asset B. That is, determine a
tan
and a
)
such that
j
c
j
= j
1
= 0.055 = target expected return
Note: The ecient portfolio will have a lower SD than asset B
Solution:
0.055 = j
c
j
= 0.03 + a
tan
(.11 .03)
a
tan
=
0.055 0.03
.11 .03
= .31
a
)
= 1 a
tan
= .69
Ecient portfolio with same expected return as asset B has
(.31)(.46) = .14 in asset A
(.31)(.54) = .17 in asset B
.69 in T-Bills
The SD of the ecient portfolio is
o
c
j
= .31(.124) = .038