Beruflich Dokumente
Kultur Dokumente
UTILITY FUNCTION
PORTFOLIO RISK AND RETURN
Reading:
Bodie, Kane and Marcus, Chapters 5, 6
Sharpe, Alexander and Bailey, Chapter 6
Utility function. Portfolio risk
and return
Definitions of:
Utility function, Diminishing marginal utility,
Diminishing marginal substitutability.
Expected utility model under uncertainty
Risk aversion, risk seeking and risk neutrality,
Indifference curves, Quadratic utility function,
Absolute and relative risk aversion.
Mean-variance (Markowitz) approach for
evaluation of risky securities
Calculation of expected return and standard
deviation of a portfolio, Covariance, Correlation
coefficient, Variance-Covariance matrix
Utility function
Utility is derived from consumption of goods and
services: level of satisfaction expressed in utils
Under certainty: no risk and no disutility
Increase in wealth, derived from investing in
assets, is a source of utility
Risk averse investors face disutility from
uncertainty of outcomes – measured through the
dispersion in expected returns
Common characteristics of utility functions
1. Nonsatiation - individuals prefer higher to lower
levels of wealth
Utility function, cont.
Common characteristics of utility functions,
2. diminishing marginal utility - as wealth increases,
utility increases but at a decreasing rate
U
Wealth (X)
22.5K
12.5K
10K
2.5K
150
100
50
50 100 150 X
U1
U2
U3
σ
Indifference curves
All portfolios that are on one indifference curve
provide the investor with the same level of utility,
although they have different expected returns and
standard deviations
Indifference curves do not intersect, as portfolios
on the same indifference curve are equivalent
Portfolios that lie on the indifference curve
furthest ‘northwest’ are the most desirable ones
There is an infinite number of indifference curves
for any investor
Indifference curves of a risk
seeking investor
Investors that choose portfolios with higher standard
deviation
E(r)
U1
U2
U3
σ
The most desirable curve is the furthest northeast
Indifference curves of a risk
neutral investor
Standard deviation is not important when evaluating
portfolios
Preferred portfolio is on the curve that is furthest north
E(r)
U1
U2
U3
σ
Risk aversion
It is assumed that all investors are risk averse
The steeper the indifference curves the higher is the degree of
risk aversion
E(r) E(r)
σ σ
Utility Function and Risk Aversion Coefficient, A
1
U E (r ) − Aσ
= 2
2
Where
U = utility
E ( r ) = expected return on the asset or portfolio
A = coefficient of risk aversion
σ2 = variance of returns
Estimating Risk Aversion
Observe individuals’ decisions when
confronted with risk
Observe how much people are willing to
pay to avoid risk
Insurance against large losses
Spreadsheet Calculations of
Indifference Curves
Indifference Curves for
U =0.05 and U =0.09 with A = 2 and A = 4
When can we use the mean - variance
approach to evaluate risky alternatives?
When there is a normal distribution of asset returns
normal distribution is bell shaped and symmetric around
the mean
mean and standard deviation perfectly describe normal
distribution
mathematical transformations may be used to convert
skewed and other non-normal distributions into a
normal one
for the right skewed distribution we add 1 to the
return and compute the natural logarithm of this
value - lognormal distribution
The Normal Distribution
Normal and Skewed Distributions
(mean = 6%, SD = 17%)
Normal and Fat-Tailed Distributions
(mean = 0.1, SD = 0.2)
Probability of Investment Outcomes After 25
Years with A Lognormal Distribution
Frequency Distributions of Rates of
Return for 1926-2005
History of Rates of Returns of Asset Classes
for Generations, US, 1926- 2005
Annually Compounded, 25-Year HPRs from
Bootstrapped History and
A Normal Distribution (50,000 Observation)
Annually Compounded, 25-Year HPRs from
Bootstrapped History(50,000 Observation)
Quadratic utility functions
When investors have quadratic utility functions
U (X) = b X + c X2
Where, U(X) is the utility from wealth, X is the wealth of
the investor and b and c are coefficients
The first derivative of quadratic utility function is showing
the marginal utility: b + 2 c X
negative c: diminishing marginal utility
The second derivative is showing the changes in marginal
utility with respect to wealth: 2c
Quadratic utility function declines after a certain level, but
the declining part is not relevant - insatiability assumption
fails
Constraint: b + 2 c X > 0
Appropriate for investments with modest returns
Absolute risk aversion
Measures how the investors preferences towards investing
in risky assets change when there is a change in wealth.
Assume that an investor has utility function U and wealth
X and a fair gamble Z whose expected value is: E(Z) = 0.
To be indifferent between the wealth and the risky gamble
on one side, and certain amount on the other side, the
investor must be paid the risk premium:
π = (1/2) σz2 [ - (U’’(X) / U’(X))]
ARA = - (U’’(X) / U’(X))
Decreasing ARA: more is invested in risky asset as wealth
increases and vice versa
Relative risk aversion
Shows the percentage change of investment in the
risky asset as wealth increases
The proportion of the risk premium would be:
p = (1/2) σz2 [ - x * (U’’(x) / U’(x))]
RRA = x * ARA = - x * (U’’(x) / U’(x))
As wealth (x) increases, the proportion of wealth
invested in the risky asset will decrease and vice
versa.
RRA can be constant
Mean-variance
(Markowitz) approach to
portfolio formation
Modern portfolio theory
Markowitz, 1952
Single period approach
Investors are Risk Averse
There is a trade off between return and risk
Asset returns follow normal distributions – therefore
can use expected returns and variances as decision
variables
Selecting an optimal portfolio from a set of portfolios
- portfolio selection problem
Maximise expected returns and minimise risk -
diversification concept
Probability Distributions
Price changes vs. Normal distribution
Microsoft - Daily % change 1990-2001
0.14
0.12
Proportion of Days
0.1
0.08
0.06
0.04
0.02
0
-9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9
Daily % Change
Probability Distributions
Standard Deviation VS. Expected Return
Investment A
20
18
16
% probability
14
12
10
8
6
4
2
0
-50 0 50
% return
Probability Distributions
Standard Deviation VS. Expected Return
Investment B
20
18
16
% probability
14
12
10
8
6
4
2
0
-50 0 50
% return
Probability Distributions
Standard Deviation VS. Expected Return
Investment C
20
18
16
% probability
14
12
10
8
6
4
2
0
-50 0 50
% return
Probability Distributions
Standard Deviation VS. Expected Return
Investment D
20
18
16
% probability
14
12
10
8
6
4
2
0
-50 0 50
% return
Theory of choice under uncertainty
The existence of risk means that there
are more than one possible outcomes
from investing in a single asset.
A probability distribution describes such
variables. For a single share example:
State of nature Probability(Pi) R1
1 0.3 10
2 0.4 20
3 0.3 30
Probability Distributions & Summary
Statistics. Example
Expected Value - Mean, E(Ri)=μ:
E(R) = Σ Ri P(Ri)
= 0.3 x 10 + 0.4 x 20 + 0.3 x 30 = 20%
Variance, V(R) = σ2:
σ2 = E{[R - E(R)]2} = E(R2) - [E(R)]2
= Σ[Ri - E(R)]2 Pi = ΣRi2 Pi - [E(R)]2
=(10-20)2 x0.3 +(20-20)2 x0.4 +(30-20)2 x0.3=60
or σ2 =(0.3 x 100 + 0.4 x 400 + 0.3 x 900) - 202 =60
Standard Deviation: σ = 7.75%
The expected return of a
portfolio of assets
Portfolio consists of a set of securities:
Rp = w1R1 + w2R2 + ... + wnRn = Σ wi Ri
Ri is the return on security i,
wi is % of portfolio value invested in the i-th security, where
Σwi = 1
Expected return of the portfolio, E(Rp), is calculated as:
E(Rp) = E[w1R1 + ... + wnRn]
= w1 E(R1) + w2 E(R2) + ... + wn E(Rn)
= Σ wi E(Ri)
Expected return of portfolio;
example
An investor has a total wealth of £10,000
and buys a portfolio with two securities.
He invests £7,000 in security A which
offers an expected return of 18%.
The remaining £3,000 are invested in
security B to produce a 12% return.
The expected return of this portfolio is:
E(Rp) = 0.7 x 18% + 0.3 x 12% = 16.2%
The expected return of a
portfolio with short selling
In the previous example, the two weights are
positive because the investor bought (is long on)
assets in the hope of selling them later at a
higher price - long position
Negative weights represent the opposite strategy
- short selling or short position
The investor borrows shares from a broker, sells
them at the current market price and buys back the
shares later (hopefully) at the lower price
Short selling incurs unlimited losses
In reality, only large financial institutions invest
proceeds from sale.
The Variance of the sum of two variables
Consider the sum of two variables: RP= R1+ R2
The variance of Rp is defined as:
σp2 = V(Rp) = E{[R-E(R)][R-E(R)]’},
where R = (R1 R2)’ is the vector of R1 and R2
σp2 = V(Rp) = V(R1+R2) = σ12 + σ22 + 2 σ12,
= σ12 + σ22 + 2 ρ12 σ1 σ2
where σ12 and σ22 are the variances of R1 and R2,
-∞<σ12<+∞ is the covariance between R1 and R2, defined
σ12=E{[R1-E(R1)][R2-E(R2)] = Σ[R1i-E(R1)][R2i-E(R2)]Pi
-1 ≤ ρ12 ≤ +1 is the correlation coefficient defined as
ρ12=σ12/σ1σ2
The Variance-Covariance Matrix
σp2 = V(Rp) = E{[R- E(R)] [R- E(R)]’}
R1 − E ( R1 ) [ R1 − E ( R1 )][ ( R2 − E ( R2 )]
= E
2R − E ( R 2
)
E{[ R1 − E ( R1 )] 2 } E{[ R1 − E ( R1 )][ R2 − E ( R2 )]}
=
E{[ R2 − E ( R2 )][ R1 − E ( R1 )]} E{[ R 2 − E ( R 2 )] 2
}
σ 1 2 σ 12
= 2
σ 21 σ 2
Calculating the Covariance
State of Nature Probability(Pi) R1 R2
1 0.3 10 -20
2 0.4 20 30
3 0.3 30 50
E(R1) = 20%, E(R2) = 21%
σ12 = E{[R1-E(R1)]2} = Σ[X1i-E(R1)]2 Pi = 60, σ22 = 789
Hence, σ1 = 7.75%, σ2 = 28.09%
Covariance:
σ12= E{[R1-E(R1)][R2-E(R2)]}=Σ[R1i-E(R1)] [R2i-E(R2)] Pi
= (10-20)(-20-21)x0.3+(20-20)(30-21)x0.4+(30-20)(50-21)x0.3
= 123 + 0 + 87 = 210
Correlation Coefficient:
ρ12 = σ12 / σ1 σ2 = 210 / (7.75 x 28.09) = 0.97
In matrix form
Expected Re turns :
E ( R1 ) 20
E ( R ) = 21
2
Variance − Co var iance Matrix :
σ 12 σ 12 60 210
=Σ = 2
σ 21 σ 2 210 789
Correlation Matrix :
1 ρ12 1 0.97
ρ =
21 1 0.97 1
The Variance of a weighted portfolio
When weights are attached to the shares of the
portfolio, that is for Rp = w1R1 + w2R2
σp2 = w12 σ12 + w22 σ22 + 2 w1w2 σ12
= w12 σ12 + w22 σ22 + 2w1w2 ρ12 σ1 σ2
w' * VC
0 0 0
1
2
3
To calculate
STOCK 4
portfolio
5
variance add
6
up the boxes
N
1 2 3 4 5 6 N
STOCK
Portfolio Risk in Matrix form
V(w’ R) = E{[w’ R – w’ E(R)] [w’ R – w’ E(R)]’}
= E{w’ [R - E(R)] w’[R - E(R)]’}
= w’ E{[R -E(R)] [R - E(R)]’} w
= w’ V(R) w w1
σ 12 σ 12 ... σ 1n .
= ( w1 ... wn ) .
2
σ n1 σ n 2 ... σ n .
w
n
=(w12σ12+...+wn2σn2) + (w1w2σ12 + w1w3σ13 + ...
+ wn-1wnσn-1,n + w2w1σ21 + w3w1σ31 + ... + wnwn-1σn,n-1)
n n n n n
σ p = ∑ wι σ ι + 2∑∑ w i w j σ i j = ∑∑ w i w j σ i j
2 2 2
i =1 i< j i =1 j =1
The variance of the portfolio
There are n variances and n(n-1) covariances
which are added together to calculate the
portfolio variance.
0
5 10 15
Number of Securities
Unique
risk
Market risk
0
5 10 15
Number of Securities
How many stocks make a diversified portfolio?
Risk Reduction of Equally Weighted Portfolios in Correlated (e.g.
shares in the same sector) and Uncorrelated (e.g. independent
insurance policies) Universes
Portfolio variance–covariance matrices for 3
shares and 10 shares
(Population) Variance-Covariance Matrix
PIST MPOK KLAO
PIST 0.0157
MPOK 0.0067 0.0399
KLAO 0.0027 0.0135 0.0374
T
Variance - Average value of squared deviations from
mean. A measure of volatility.
T
∑ t
( R − R ) 2
σˆ 2 = t =1
T −1
Standard Deviation – Square root of variance;
volatility in units of measurement of variable.
T
Covariance:
∑ (R 1t − R1 )( R2t − R2 )
σˆ 12 = t =1
T −1
Mean return & Variance-covariance matrix of returns for
3 ASE stocks. Period: 1990:1- 2001:6
Boutaris 2.00%
(MPOK), 1.50%
1.62%,PISTKLAO
1.35%,PISTMPOK
Klaoudatos 1.00%
0.66%, KLAO
(KLAO) 0.50% 0.38%,MPOKKLAO
0.11%, MPOK
0.00%
0.0000 0.0500 0.1000 0.1500 0.2000 0.2500
Standard Deviation
Earlier portfolios were set with arbitrary weights
(50% in each stock).
The benefits of diversification are obvious, as we
move northwesterly in relation to holding single
shares.
When weights are determined which minimize
the variance s.t. sum of weights =1 then the
global minimum portfolio is determined.
When variance is minimized s.t. sum of
weights=1 over a range of required rates of
return on the portfolio, then the efficient frontier
is determined.
We consider the last two problems later on.