Professional Documents
Culture Documents
𝜇$ = , 𝑤% 𝜇% = 𝑤 ' 𝜇
%&!
# #
𝑤 # Σ𝑤 = ∑" "
$%! ∑'%! 𝑤$ 𝑤' 𝜌𝑖𝑗𝜎$ 𝜎' (2)
§ Assumptions (important!)
§ Single period model
§ Assets are perfectly divisible
§ No transaction costs and no taxes
§ Maximize return, minimize variance, or both
§ Mean returns 𝜇 and covariance Σ are known
§ The variance-covariance matrix 𝛴 is positive definite
§ Not all expected returns are equal (𝜇 is not a multiple of 𝑒)
Mean-Variance Portfolio Optimization – Formulations
(
§ Maximize portfolio return for given risk tolerance, 𝜎*+,
max 𝑤 # 𝜇 𝑠. 𝑡. 𝑤 # Σ𝑤 = 𝜎*+,
( , 𝑒#𝑤 = 1
-
§ The Lagrangian
2
𝐿 𝑥, 𝜆 = f x − Q 𝜆1 (𝑔1 𝑥 − 𝑐1 )
1%!
§ Optimality conditions
𝜕𝐿
= 0, ∀𝑖 = 1, ⋯ , 𝑁 ⇔ ∇, 𝐿 = 0
𝜕𝑥$
𝜕𝐿
= 0, ∀𝑘 = 1, ⋯ , 𝐾 ⇔ ∇3 𝐿 = 0
𝜕𝜆1
Use Lagrange multiplier to maximize risk-adjusted return
1
𝐿 𝑤, 𝜆 = 𝜏 ⋅ 𝑤#𝜇 − ⋅ 𝑤 # Σ𝑤 − λ(𝑒 # 𝑤 − 1)
2
∇- 𝐿 𝑤, 𝜆 = 𝜏 ⋅ 𝜇 - Σ𝑤 – λ ⋅ 𝑒=0
∇7 𝐿 𝑤, 𝜆 = -(𝑒 # 𝑤 − 1) =0
Use Lagrange multiplier to maximize risk-adjusted return
§ ∇- 𝐿 𝑤, 𝜆 = 𝜏 ⋅ 𝜇 - Σ𝑤 - λ ⋅ 𝑒=0 ⟹ 𝑤 = 𝜏 ⋅ Σ 8! 𝜇 - λ ⋅ Σ 8! 𝑒 (1)
§ (2) → (1)
9 # :$! ; !
𝑤456 = 𝜏 ⋅ Σ 8! 𝜇 – (𝜏 ⋅ # $! − # $! ) ⋅ Σ 8! 𝑒
9 : 9 9 : 9
# $!
9 : ; $!
: 9
= 𝜏 ⋅ (Σ 8! 𝜇 – # $! Σ 8! 𝑒) + # $!
9 : 9 9 : 9
9 # :$! ; 8! :$! 9
§ 𝑤456 = 𝜏 ⋅ (Σ 8! 𝜇 – # $! Σ 𝑒) + # $! =𝜏 ⋅ 𝑤< + 𝑤*,
9 : 9 9 : 9
# $!
9 : ; 8! :$! 9
where 𝑤< = Σ 8! 𝜇 – # $! Σ 𝑒 and 𝑤* = # $!
9 : 9 9 : 9
Use Lagrange multiplier to maximize risk-adjusted return
§ 𝑤456 = 𝜏 ⋅ 𝑤< + 𝑤*
( = 𝑤# Σ 𝑤
§ 𝜎456 ( # # #
456 456 = 𝜏 ⋅ 𝑤< Σ𝑤< + 2𝜏 ⋅ 𝑤* Σ𝑤< + 𝑤* Σ𝑤*
=𝜏 ( 𝜎=( + 2𝜏𝜎*< + 𝜎*(
§ Prove (HW)
𝜎*< =0
𝜇< = 𝜎=(
𝑒 # 𝑤< =0
Optimal Risk-Reward Trade-off
§ If 𝑃 is any optimal portfolio 𝑤456 then
𝜇5 = 𝜇* + 𝜏𝜇< , 𝜎5( = 𝜎*
( + 2𝜏𝜎 ( (
*< + 𝜏 𝜎<
§ Provided 𝜎*< = 0 and 𝜇< = 𝜎<( , we have
;% 8;&
𝜇5 = 𝜇* + 𝜏𝜇< ⟹ 𝜏 = (1)
;'
>%( 8>&
(
𝜎5( = 𝜎*
( + 𝜏 (𝜎 ( ⟹ 𝜏 ( =
< (2)
>'(
>%( 8>&
( ;% 8;&
(1) → (2) =( )( (3)
>'( ;'
(;% 8;& )(
𝜇< = 𝜎<( ⟹ 𝜎5( − (
𝜎* = (4)
;'
Parabola in (𝜎 ( , 𝜇)
Hyperbola in (𝜎, 𝜇)
Efficient Frontier
𝜇/ ntie
r
Fr o
ient
ic
Eff
𝜇0 Min risk
portfolio
Individual Assets
sm sP
Two-Fund Theorem
§ 𝑤456 = 𝜏 ⋅ 𝑤< + 𝑤*
§ As long as you can identify 2 efficient portfolios, you can obtain any
efficient portfolios (two-fund theorem)
Summary: Portfolio Optimization with Only Risky Assets
For a fixed number 𝜏 ≥ 0, the optimal solution for
1
max 𝜏 ⋅ 𝑤 𝜇 − ⋅ 𝑤 # Σ𝑤
# 𝑠. 𝑡. 𝑒#𝑤 = 1
- 2
is 𝑤456 = 𝑤* + 𝜏𝑤< , where
§ 𝑤* is the min-risk portfolio and 𝑤< is a “zero-covariance” portfolio
§ 𝜎*< = 𝑤* # Σ𝑤 = 0, hence the name “zero-covariance”
<
#
§ 𝑒 𝑤< = 0, 𝑤< is a self-financing portfolio
§ 𝑒 # 𝑤* = 1, the min-risk portfolio satisfies the budget constraint
§ Graphical identification
§ Efficient frontier & the risky feasible set with only risky assets
§ The risky asset 𝑄 can be any point within the feasible set
§ Graph the resulting combined portfolio 𝑃 in the 𝜎, 𝜇 -plane
§ Identify the portfolios with the best risk-reward tradeoff
1 Risky Asset + 1 Risk-Free Asset
+&
§ Substitute 𝑤) = 1 − into the first equation, we have
+'
𝜇2 − 𝑟3
𝜇/ = 𝑟3 + 𝜎/
𝜎2
,' -.(
§ Straight line with intercept 𝑟* & slope
+'
§ Known as the Capital Allocation Line (CAL)
§ Different risky assets 𝑄 result in CALs with different slopes
,' -.(
§ Slope measures the excess return over 𝑟* per unit of risk
+'
Capital Allocation Line
;) 8C*
§ CAL: 𝜇& = 𝑟B + 𝜎&
>)
§ 𝑄 can be any portfolio in the feasible set 𝜇 Q’
+
L
M
§ Higher slope is better C
sP
One-Fund Theorem
§ There is a unique optimal risky portfolio M such that any efficient
portfolio can be constructed as a combination of M & risk-free asset 𝑟B
§ This is called the one-fund theorem
§ The new efficient frontier is a line, the CML
§ What is the new feasible set?
𝜇)
L
CM
Tangent
Portfolio
rf
sP
Effect of risk-free rate 𝑟c
1
L
CM
è slope of CML decreases
è 𝜎D & 𝜇D increase
§ Whatever 𝑟B is 𝑟B(
Second Optimal
Risky Portfolio
§ CML remains a straight line First Optimal
Risky Portfolio
§ CML remains the steepest CAL
§ Tangent portfolio shifts accordingly
𝑟B!
sP
The Separation Principle
§ Portfolio selection can be separated into two stages:
1. Determine the tangent portfolio M
§ Need 𝜇 & Σ to find the risky efficient frontier and 𝑟B to identify M
§ Homogeneous expectations: all investors share the same 𝜇 and Σ
Ø All investors who believe the same 𝜇 and Σ have the same M
Ø Portfolio M does not depend on investors’ risk tolerance
2. Decide the optimal mix of tangent portfolio and the risk-free asset
§ This depends on investor’s risk tolerances (𝜏 in our formulation)
§ It also assumes all investors care about expected return and variance
§ Almost always false
§ Some investor may care skewness, kurtosis, etc.
§ Some investor may care about variance in losses only
!
§ Consider the “ -portfolio” for 𝑁 assets with common variance 𝜎A( and
"
common covariance 𝜌𝜎A(
, , #, ,
# ,
# 𝜎- 𝜌𝜎- 1 # 1
𝜎) = 5 5 𝑤! 𝑤" 𝜎!" = 5 # + 5 5 # = 𝜎- + 1 − 𝜌𝜎-#
𝑁 𝑁 𝑁 𝑁
!*+ "*+ !*+ !*+ "*+
".!
§ Diversify the portfolio by 𝑁 → ∞, then 𝜎&( ≈ 𝜌𝜎A(
𝑵 1 2 5 10 100 500 ∞
𝝆=𝟎 1 0.5 0.2 0.1 0.01 0.002 0
𝝆 = 𝟎. 𝟑 1 0.65 0.44 0.37 0.307 0.301 0.300
Types of Risk
§ The total risk of a security can come from two sources:
§ Systematic risk
§ a.k.a. market risk, or non-diversifiable risk
§ Macro risk factors that affect essentially all assets.
(Although not necessarily all to the same degree)
§ Examples: coronavirus, financial crises, political instability
§ Non-systematic risk
§ a.k.a. specific risk, diversifiable risk, idiosyncratic risk
§ Micro risks that affect essentially one security only.
Portfolio risk
Idiosyncratic risk
Systematic risk
Level of diversification
Summary
§ Portfolio optimization
§ Risky assets only: the efficient frontier is a hyperbola
§ Inclusion of risk-free asset:
§ Capital Market Line (CML) is the efficient frontier, a straight line
§ Capital Allocation Line (CAL): the steepest one is the CML
§ Separation theorem:
1. First find the risky efficient frontier (same for every investors)
2. Mix portfolio M with 𝑟B (different mix for different investors)
§ Risk diversification
§ Benefit: portfolio risk less than sum of individual risks
§ Limitation: systematic risk is not diversifiable, due to correlation