Model Portfolio Methodologies

Mean Variance optimization process (MPT / Markowitz Bullet)

Develops optimization problem based on constraint of maximizing expected return 1
and minimizing risk
Advantages
1. Academically sound
Issues
1. Sometimes produces returns that are extreme and not intuitive
2. User have to provide a complete set of expected returns for generating
optimal portfolio weight
3. Model risk high due to requirement of absolute values of expected returns
and portfolio weights
4. Markowitz formulation requires expected returns for all components of the
relevant universe which is very exhaustive as PMs tend to focus on small
segments like picking stocks, stocks with positive momentum and identifying
relative value trades
5. Results can be extreme if PMs focus in asset weights and ignore expected
returns

Black Litterman model

Combines the market equilibrium with additional market views of the user
With inputs in form of views or statements about the expected returns, the model
combines the views with equilibrium producing the set of expected returns of asset
as well as the optimal portfolio weights. It is simply a set of deviations from market
capitalization weights in the direction of portfolios about the expressed views.
Market Equilibrium: Reference Point for Black Litterman Model
The Black-Litterman model starts with equilibrium expected returns. According to
the Capital Asset Pricing Model (CAPM), prices will adjust until the expected returns
of all assets in equilibrium are such that if all investors hold the same belief, the
demand for these assets will exactly equal the outstanding supply. This set of
expected returns is the neutral reference point of the Black-Litterman model. The
investor then can express her views about the markets.
In the Black-Litterman model, a view is a general statement about the expected
return for any portfolio. These views are combined with the market equilibrium
expected returns. In the case when the investor does not have any views about the
1 Expected return means expected excess return over the risk free yield curve

markets, the expected returns from the Black-Litterman model match the
equilibrium, and the unconstrained optimal portfolio is the market equilibrium
(capitalization weights) portfolio. In the case when the investor has one or more
views about the market, the Black-Litterman approach combines the information
from the equilibrium and tilts the optimal portfolio away from the market portfolio in
the direction of the investors views.
Constraints
Expected Returns (multiple views of relative outperformance) In general
portfolio weight increases as expected returns increases
Degree of confidence
The constrained Optimal portfolio BL not very intuitive in these cases
Risk Constraint weights are adjusted by scaling target risk levels with
volatility of unconstrained optimal portfolio
Budget Constraint
Beta Constraint
Advantages
1. Intuitive
2. Systematic approach to demonstrate deviation of users market view from
reference equilibrium
3. Market Capitalization portfolio with equilibrium expected returns provide
optimal portfolios on which views can be superimposed easily
4. Selected views on regions, sector outperformance etc can be easily
incorporated
5. Simplistic way of incorporating views on portfolios using Markowitz
framework rather than working on complete vector of expected returns.
Benefits
1. The unconstrained optimal portfolio is the market equilibrium portfolio plus a
weighted sum of portfolios calibrated with users view
2. As expected the weight of an asset in portfolio increases with positive view in
comparison to one implied by equilibrium
Calculation Procedure
1. There are
portfolio) is
returns:

assets in the market. The market portfolio (equilibrium

w eq . The covariance of the returns is

. The expected

is a vector of normally distributed random variables with mean

2. The average risk tolerance of the world is represented by the risk-aversion

parameter. The equilibrium expected returns are

= w eq . The CAPM

= +

prior distribution for the expected returns is

normally distributed with mean zero and covariance

(e)

, where

(e)

is

. The parameter

is a scalar measuring the uncertainty of the CAPM prior.

3. The user has

K N

is a

Q+ (v)

views about the market, expressed as

matrix and

is

(v)

-vector, and

distributed with mean zero and covariance

, where

is normally

. The users views are

independent of the CAPM prior and independent of each other.

4. The mean of the expected returns is

( )
[ 1 + P ' 1 Q ].
1

=[ ( )1+ P' P ]
1

5. The investor has the world average risk tolerance. The objective of the
investor is to maximize the utility
optimal portfolio is

= 1 /

' w ' w /2 . The unconstrained

, which can be written as

w =weq + P '

Since the columns of matrix P are the portfolios in the users view, this
means that the unconstrained optimal portfolio is the market portfolio plus a
weighted sum of the portfolios in the users views. The weights for these
portfolios are given by the elements of the vector
1

formula
6. Let

, which is given by the

1

1 Q / [ / + P P' ] P weq [ /+ P P' ] PP ' 1 Q/ .

P ,Q

and

represent the

views held by the investor initially,

be the expected returns by using these views in the Black-Litterman

model,

be the weight vector defined above. Assume the investor now

has one additional view, represented by

K +1 views, the new weight vector

p,q

and

. For the new case of

is given by the following formula'

where A = + PP, b = Pp, c = + pp. Since c bA1b > 0 , the

expression of shows the additional view will have a positive (negative)
weight if q p > 0 ( q p < 0 ), this corresponds to the case where the
new view on the portfolio p is more bullish (bearish) than implied by the old
expected return . The additional view will have a zero weight if q = p , this
corresponds to the case where the new view is implied by the old expected
returns already. In this case, the new view has no impact at all.

7. For a particular view k , its weight k is an increasing function of its expected

return qk . The absolute value of k is an increasing function of its confidence
level k 1.
Solution for constraints
1. Given the expected returns and the covariance matrix , the unconstrained
maximization problem max w w ww 2 has a solution of w* = ( ) 1
.
2. Given the covariance matrix , the minimum variance portfolio is w(m) =
1 (1), where is a vector with all elements being one.
3. The solution to the risk constrained optimization problem, max w , subject
to ww 2 , can be expressed as * * * ( ) w w w w r = , where w* = ( )
1 is the solution of the unconstrained problem.
4. The risk and budget constrained optimization problem can be formulated as
max w , subject to ww 2 and w = 1 . Its solution has the form w(b) =
aw*+bw(m) , where a and b are chosen in the way both risk and budget
constraints are satisfied.
5. The risk-, budget-, and beta-constrained optimization problem can be
formulated as max w , subject to ww 2 , w = 1 , and wweq = we
qweq , where weq is the market portfolio. The solution to the problem has
the form of w aw bw m cw eq ( ) = *+ ( ) + , where a , b , and c are chosen
in the way all three constraints are satisfied.
Issues
Criticism