AP I - 2. Portfolio Choice Mean Variance Approach

Univ.-Prof. Thomas Gehrig, Ph.D.
Asset Pricing I, Winter 2019/20
Asset Pricing I
Winter 2019/20
2. Portfolio Choice: Mean Variance Approach
1
Univ.-Prof. Thomas Gehrig, Ph.D. Asset Pricing I, Winter 2019/20
Basic idea: Trade-off between risk and return can easily be analyzed graphically in a socalled
(µ − σ )-diagram. Any security is characterized by its first and second moments.
2
Source: Chicago Center for Research in Security Prices (CRSP) for U.S. stocks and CPI, Global Finance Data for the World Index, Treasury bills
and corporate bonds.
3
Return Distributions
Graphs in the sequel are taken from Berk, de Marzo: Corporate Finance, 2nd edition, 2011, chapters 10-11.
4
Historical Tradeoffs between Return and Risk
5
Historical Volatility and Return for 500 Individual Stocks

USA: 1926-2005 quarterly data
Berk, de Marzo: Corporate Finance, 2nd edition, 2011, chapters 10-11.
6
Risk Diversification (1/2)
Consider a portfolio with a share λ ∈ [0, 1] of the risky security x̃ and 1 − λ of security ỹ.
The portfolio characteristics are
• mean µλ = λ µx + (1 − λ )µy
• variance σλ2 = λ 2σx2 + (1 − λ )2 σy2 + 2λ (1 − λ )σxy
â Hence portfolio risk is decreasing in the covariance σxy

â Diversification potential increases as the covariance σxy decreases
Note: There is NO diversification potential with a riskless asset, since in this case σxy = 0.
7
Risk Diversification (2/2)
The role of correlation:
σxy covxy
Def.: Correlation coefficient corrxy = =
σx σy sdx sdy
Note: The correlation coefficient is normed between −1 ≤ corrxy ≤ 1
â If corrxy = 1 σλ2 = λ 2σx2 + (1 − λ )2 σy2 + 2λ (1 − λ )σx σy

Hence: σλ = λ σx + (1 − λ )σy
â If corrxy = −1 σλ2 = λ 2σx2 + (1 − λ )2 σy2 − 2λ (1 − λ )σx σy

Hence: σλ = |λ σx − (1 − λ )σy|
8
Correlation and Risk Diversification
Definition: Diversification is the sub-linear relation in the risk of constituting elements of a portfolio.
9
Tradeoff between Volatility and Expected Return
10
Allowing for Short Positions
11
Three securities (1/2)
12
Three securities (2/2)
13
Moving towards the efficient frontier of the (full) market
14
Introducing a risk free asset
15
Minimum Variance Opportunity Set and Tangency Portfolio
Which conditions are needed for the minimum variance opportunity set (=efficient frontier) to be
well-behaved?
Lemma 3.1: If there are finitely many assets, the minimum-variance opportunity set is closed and
connected.
Result 3.2: If there are finitely many assets and at least one asset has a mean return exceeding
the riskfree return R0, then a tangency portfolio exists.
Corollary 3.3: If there are finitely many assets a tangency portfolio exists whenever shortselling is
feasible.
16
The Capital Market Line
17
Two Fund Separation (Tobin 1958)
Result 3.4: For any (mean-variance) preference the optimal portfolio choice consists of a port-
folio of the riskfree asset and the tangency portfolio.
18
Computing the Tangency Portfolio
Example: Consider a risk free security x0 and K risky securities x˜k , k = 1, ..., K
bi 2 K
i
µλ , σλ2

Let U = µλ − σλ , where the portfolio shares add up to 1, ∑ λk = 1.
2 k=0
i 0 bi 0
Optimal portfolio shares: λ̂ ∈ argmax(µ − R0ı) λ − λ COV λ , i = 1, ..., K
2
1
Solution with short sales: λ̂ i = COV −1 i (µ − R0ı)
b
λ̂k
λk = K
∑ λ̂l
l=0
19
Market Equilibrium (1/6)
Asset demand: Let the prices of securities k = 1, ..., K be denoted by pk . Security supply is zk .
Initial wealth: W0i ∈ ℜ ≥ 0
Portfolio: θ ∈ ∆K ⊂ ℜK
Budget constraint: θ0 + θ p = W0i
i 0 bi 0
Optimal portfolio choice: θ̂ ∈ argmax(µ − R0 p) θ − θ COV θ
2
1
Demand for risky securities: θ̂ i = COV −1 i (µ − R0 p)
b
Discussion: Comparative statics?

Which properties are less plausible?
20
 
 
1  1 
Market Equilibrium: p=  µ − I
COV ∗ z
R0 
 1 

∑ bi
i=1
Due to two-funds separation each investor optimally invests in a combination of the riskless asset and
the tangency portfolio.
Hence in equilibrium each investor holds a mixture of the tangency portfolio and the riskless security.
Accordingly, the tangency portfolio is identical to the market portfolio. I.e. equilibrium prices will
adjust such that in equilibrium each investors optimally holds the tangency portfolio, which then is the
market portfolio.
21
Capital Market Line (CML)
22

d M

µ λ R0 + (1 − λ )R
dλ λ =0 R0 − µ M
Slope of Capital Market Line: =
d −σM
σ λ R0 + (1 − λ )RM

dλ λ =0

d M

µ λ R j + (1 − λ )R
dλ =0 µj − µM
Slope of Portfolio x˜j , x̃M :
λ
= M 2

d COV R j , R − σ
σ λ R j + (1 − λ )RM
M
dλ λ =0 σM
M

COV xk , x
µk − R0 = βk µ M − R0 , where βk =

Tangency condition yields:
σM2
23
Security Market Line (SML): Sensitivity with respect to systematic risk as measured by βk .
24
Empirical implications of market equilibrium provide a basis for testing the theory:
M
!
• Pricing implications SML µk − R0 = αk + βk µ − R0 → Test αk = 0
Fama-French factors: refinements or anomalies?
â Value
â Size
â Momentum
â Beta
• Quantitiy implications θi = θ M
Contradictions: Home bias
⇒ What is wrong about the model? It is too simple!
E.g. µ 1 = ... = µ i = ... = µ (same for higher moments)
25
Market Equilibrium with Heterogeneous Beliefs (1/2)
Let us now allow for heterogeneous expectations (first moments) but keep all the other moments -
especially the risk assessments - identical.
1
Optimal individual portfolio choice: θ̂ i = COV −1 i (µ − R0ı)
b
Theorem: In the CAPM with heterogeneous beliefs the SML holds for the average beliefs, i.e.
I
for all assets k = 1, ..., K we have µ̄k − R0 = βk (µ̄ − R0), where µ̄ = ∑ ai µ i is
M M
i=1
i
r
i b i
i wi0 i i
i
the average belief with weights a = I j and r = I with w0 = 1 − θ0 W0 ,
r
∑ bj ∑ w0j
j=1 j=1
where W0i is the initial wealth of investor i.
26
Market Equilibrium with Heterogeneous Beliefs (2/2)
Implications:
• Two fund separation fails

• Each investor has his private security line
• Quantity implication does no longer hold
• Beta remains unchanged (Remember: Second moments were assumed to be identical across in-
vestors!)
• Alpha is investor specific - some investors may have positive and others negative alpha
• Tangency and market portfolio are no longer the same
• It makes sense to distinguish active from passive trading strategies: tangency portfolio (active)
versus market portfolio (passive)
• The average market belief depends on wealth and risk aversion
â Stronger influence of wealthier investors
â Stronger influence of less risk averse investors
27
Active versus Passive Fund Management (1/2)
Active fund management: Investment in individual tangency portfolio requires costly re-
sources (e.g. information)
Passive fund management: Investment in market portfolio - requires no extra resources
1
Active investor’s beliefs: λ̂ i = COV −1 (µ − R0ı)
bi
µ i − R0

1
µ̄ki − R0 = βk µ̄ M − R0 we get λ̂ i = i λM

If in equilibrium: 2
b σM
Does this imply that active management is irrelevant?
28
Active versus Passive Fund Management (2/2)
Answer: No! Need to compare utility of active management with outside option of passive
management.
i i 2 bi 2
λ0i R0 + 1 − λ0i i
1 − λ0i VAR Rλ0i −Ci

Utility of active investor: U µ ,σ = µ̂ Rλ −
2
i
2 b
Utility of passive investor: i
U (µ̄, σ ) = λ̄ i
R0 + (1 − λ̄0i )µ̂(Rλ̄ ) − (1 − λ0i )2VAR(Rλ̄ )
2
Which of the two dominates? Active utility is more likely to be larger
• the more skilled the active investor

• the smaller the cost of active management
• the less risk averse the investor
29
Richer Pricing Models
• Alternative (additional) Betas: Include more risk factors (e.g. background risk, liquidity risk)
• Higher Moment Betas: Include more information about the return distribution (e.g. skewness and
catastrophic risk)
• Behavioral CAPM: Include asymmetric behavioral biases of PT and/or CPT
30
Alternative Betas
Additional factors not modeled in the simple model may affect pricing. Such factors can be
• background risk
• liquidity risk (see LAPM in the last section of this course)
F
Like Abitrage Pricing Theory: RM = ∑ γfRf
f =1
F
Security market line: µk − R0 = ∑ βk f (µ f − R f )
f =1
γ f COV (Rk , R f )
where βk f =
σM2
31
Higher Moment Beta - CAPM
Extend mean variance preferences to allow for higher moments.
• Catastrophic risk (left skewed distributions)

• "fat tails"
Numerical optimization.
32
Behavioral CAPM

u(c), if c > RP
Asymmetric utility: v(c) = 1
− u(c), if c < RP
β

α+
∆x − (∆x)2,

 if ∆x > 0
where u(∆x) = 2 and ∆x = x − RP
α−

2
β ∆x − (∆x) , if ∆x < 0


2
Prospect gains: pt +(c) = ∑ πsv(cs)

cs >RP
Prospect losses: pt −(c) = ∑ πsv(cs)

cs <RP
Overall prospect utility: PTu(c) = pt +(c) − β pt −(c)
33
Similar approach as before in the gains-losses diagram.
(µ − σ )-approach can be viewed as a special case of the general gains-losses approach.
34
Result:
• Gains and losses are determined according to the returns of the market portfolio relative to the
reference point!
• Gains and losses are determined according to the gradient of the value function
• Hence, only individual recommendations.
• No recommendations for the holder of the market portfolio.
35

AP I - 2. Portfolio Choice Mean Variance Approach

Uploaded by

Copyright:

Available Formats

AP I - 2. Portfolio Choice Mean Variance Approach

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

AP I - 2. Portfolio Choice Mean Variance Approach

Uploaded by

Copyright:

Available Formats

Univ.-Prof. Thomas Gehrig, Ph.D.

Asset Pricing I, Winter 2019/20

2. Portfolio Choice: Mean Variance Approach

Historical Tradeoffs between Return and Risk

Historical Volatility and Return for 500 Individual Stocks

Berk, de Marzo: Corporate Finance, 2nd edition, 2011, chapters 10-11.

Risk Diversification (1/2)

• variance σλ2 = λ 2σx2 + (1 − λ )2 σy2 + 2λ (1 − λ )σxy

â Hence portfolio risk is decreasing in the covariance σxy

Risk Diversification (2/2)

The role of correlation:

Note: The correlation coefficient is normed between −1 ≤ corrxy ≤ 1

â If corrxy = 1 σλ2 = λ 2σx2 + (1 − λ )2 σy2 + 2λ (1 − λ )σx σy

â If corrxy = −1 σλ2 = λ 2σx2 + (1 − λ )2 σy2 − 2λ (1 − λ )σx σy

Correlation and Risk Diversification

Tradeoff between Volatility and Expected Return

Berk, de Marzo: Corporate Finance, 2nd edition, 2011, chapters 10-11.

Allowing for Short Positions

Three securities (1/2)

Three securities (2/2)

Moving towards the efficient frontier of the (full) market

Berk, de Marzo: Corporate Finance, 2nd edition, 2011, chapters 10-11.

Introducing a risk free asset

Berk, de Marzo: Corporate Finance, 2nd edition, 2011, chapters 10-11.

Minimum Variance Opportunity Set and Tangency Portfolio

The Capital Market Line

Berk, de Marzo: Corporate Finance, 2nd edition, 2011, chapters 10-11.

Two Fund Separation (Tobin 1958)

Computing the Tangency Portfolio

Market Equilibrium (1/6)

Discussion: Comparative statics?

Market Equilibrium (2/6)

Market Equilibrium (3/6)

Capital Market Line (CML)

Market Equilibrium (4/6)

Market Equilibrium (5/6)

Market Equilibrium (6/6)

Market Equilibrium with Heterogeneous Beliefs (1/2)

Market Equilibrium with Heterogeneous Beliefs (2/2)

• Two fund separation fails

Active versus Passive Fund Management (1/2)

Passive fund management: Investment in market portfolio - requires no extra resources

Does this imply that active management is irrelevant?

Active versus Passive Fund Management (2/2)

Which of the two dominates? Active utility is more likely to be larger

• the more skilled the active investor

Richer Pricing Models

• Behavioral CAPM: Include asymmetric behavioral biases of PT and/or CPT

Higher Moment Beta - CAPM

Extend mean variance preferences to allow for higher moments.

• Catastrophic risk (left skewed distributions)

Prospect gains: pt +(c) = ∑ πsv(cs)

Prospect losses: pt −(c) = ∑ πsv(cs)

Overall prospect utility: PTu(c) = pt +(c) − β pt −(c)

Similar approach as before in the gains-losses diagram.

(µ − σ )-approach can be viewed as a special case of the general gains-losses approach.

You might also like