Asset Allocation II
MGMT 645: Portfolio Management
Recap and agenda
• Last class, we found optimal portfolios of
• A risk-free and a risky asset
• A risk-free and two risky assets
• Now … wait for it …
• A risk-free and three risky assets
• A risk-free and more risky assets
• Main issue: maximizing diversification benefit
Adding a third risky asset
• Let’s add gold (GLD) to the menu of stocks and
corporate bonds Should be
VTI VCLT GLD some
m 13% 7% 2% diversification
s 18% 12% 15% benefit
Correlations VTI VCLT GLD
VTI 1.00 0.21 0.03
VCLT 0.21 1.00 0.29
GLD 0.03 0.29 1.00
m=0.08
s=0.10
w=0.32, 0.50, 0.18
Gold hedges m=0.08
s=0.11
w=0.17, 0.83
volatility
Frontier
dominates
individual
assets
How the problem changes
• Before, we had the constraint
• Therefore, it had to be that
• Choose and calculate the portfolio mean and
variance
• Repeat for different to trace the frontier
Optimization
• Now we have to find , , and to minimize variance for
a given expected return
• Formally (in words)
Find weights , , and to minimize portfolio variance, ,
given a desired expected return, , with the constraint
that the weights have to sum to one.
Three risky assets: expected returns
and variance
• Expected return adds one new term
• Several new terms in the variance
How the problem changes
• Now, we have the constraint
• It has to be that
• Now if I choose , many possible and
• Need to check all possible values to get the best risk-
return tradeoff
Implementation
1. Set up a column of weights in Excel
2. Set up a cell with the sum of the weights
3. Set up a cell with the weight-dependent standard deviation
4. Set up a cell with the weight-dependent expected return
5. Set up a cell with the target expected return
6. Use Solver to
a. Minimize the cell in step 3
b. By changing the column of weights in 1
c. Subject to the constraints that
i. The cell in 2 is equal to 1
ii. The cell in 4 is equal to the cell in 5
7. Repeat for different values of #5 to trace the frontier
Practice: International Diversification
• VTI: U.S. All-stock index
• VEA: Developed ex-US stock index
• VWO: Emerging stock index
• Efficient frontier of VTI and VEA vs VTI, VEA, and VWO
A useful shortcut
• Trace out frontier: solve over and over again for
different m*
• Fisher Black showed that any efficient portfolio is a
portfolio of any two other efficient portfolios
• If portfolios u and v are efficient, a portfolio auwu + (1-au)wv
is also efficient
• Implication: don’t have to solve for every point on the
efficient frontier – just have to solve for two sets of portfolio
weights
The minimum variance portfolio
• Two portfolios of particular interest:
• The minimum variance portfolio
The maximum Sharpe portfolio
• Two portfolios of particular interest
• The maximum Sharpe ratio portfolio
Covariance of MV and MS portfolios
• To use this, you need the covariance of the minimum variance (MV) and
maximum Sharpe (MS) portfolios
• Let
then
Practice: Black
Generalizing: portfolio mean
• Mean return for a portfolio of N assets:
• More compactly,
Generalizing: portfolio volatility
• Volatility is much uglier:
• More compact in linear algebra
• I’ll set things up in spreadsheets for you
• If anyone is interested, linear algebra explanations at the
end
The general problem
Restricting short sales
• Simply click the Solver box “Make unconstrained
values nonnegative”
• Drawback: can’t use the Black shortcut anymore
Maximum Sharpe and CAL
• In the case with two risky assets, we found a CAL by
finding the line tangent to the efficient frontier with
intercept = Rf
• Recall that the slope was
where T designates the tangency portfolio
Bringing back the
risk-free asset
• Notice that this is the line
with maximum slope that
goes through both the
efficient frontier and the
risk-free rate
• That is, it is the portfolio
that maximizes the Sharpe
ratio
Steeper CAL =
Increased Utility
• The investor can now increase utility
• Indifference curve shifts northwest
Benefit of adding
assets
• Adding GLD expands the
frontier
• It also steepens the CAL
• Tangency Portfolio with
better risk-reward ratio
• Results in yet another
better opportunity set
• CAL portfolios dominate
the efficient frontier
Practice: Optimal portfolios
Realistic constraints
• No short sales
• Many institutions are prohibited from shorting
• Shorting is costly and can be hard to execute (see Gamestop)
• Implemented by setting unconstrained variables (weights) to
nonnegative
• No borrowing
• Before, single risky asset portfolio = index
• Harder to find levered version of tangent portfolio
Constrained frontier and optimal
portfolios
Concluding remarks
• Asset allocation is a two-step process
• Find a risky asset portfolio that maximizes the reward-risk
tradeoff
• Balance with cash to accommodate risk aversion
• Real-world constraints can limit the opportunity set
• Short selling
• Borrowing
Appendix: Mean-Variance in
Linear Algebra
Vectors and transposes
• A column vector is an column-oriented array of
numbers,
• Its transpose is a row-oriented array (a row vector)
Matrices
• Matrices are 2-dimensional arrays of numbers,
Matrix multiplication
• Multiplying two vectors
• Multiplying a matrix and a vector
Portfolio moments
• Portfolio mean
• Portfolio variance
Matrix multiplication in Excel
CTRL+SHIFT+ENTER to enter an array formula like mmult
Matrix Multiplication in Excel
Mean-variance problems in linear
algebra
• General portfolio optimization
Mean-variance problems in linear
algebra
• Minimum variance portfolio
Mean-variance problems in linear
algebra
• Maximum Sharpe Ratio portfolio
Appendix: Pictorial
Illustration of Optimization
Step 1: Set up column of weights
Step 2: Cell with sum of weights
Step 3: Cell with standard deviation
Step 4: Cell with expected return
Step 5: Cell
with target
return
Step 6:
Solver
Solver
weights
Step 7: Repeat
Efficient weights
Efficient
Frontier
