Jessica Wachter Class notes for Finance 604

Topic 8. Portfolio Analysis

(a) Two Risky Assets

Last time, we calculated individual assets’ means and standard deviations. However, as
investors, we don’t just choose individual assets - we can combine them to form
portfolios.
Consider 2 assets: how does our portfolio return relate that of our underlying assets?
Notation:

• X1 = Portfolio weight in asset 1

• X2 = Portfolio weight in asset 2

• R1 = Return on asset 1

• R2 = Return on asset 2

• Rp = Return on portfolio.

Result:
Rp = X1 R1 + X2 R2 .

⇒ This is the expected return of a two-asset portfolio. We will prove this with an
example.

Example Assume that we have 2 securities and $100 ⇒ we put $20 in 1 and $80 in 2.
Thus, the portfolio weights would be:
20
X1 = = 0.2.
100
80
X2 = = 0.8.
100

What’s true about these weights?

X1 + X2 = 1.

All your money must go to something. Also, you can’t invest money you don’t have.

1
Jessica Wachter Class notes for Finance 604

How much did you earn on your portfolio? We calculate holding period return:

$20(1 + R1 ) + $80(1 + R2 )
Rp = − 1.
$100
= 0.2(1 + R1 ) + 0.8(1 + R2 ) − 1.
= 0.2R1 + 0.8R2

Note that this agrees with our result above.

Of course, the given R1 and R2 represent only one of a set of possible outcomes. Recall
the example from last time, with three scenarios, each with equal probability:

Scenario R1 R2 Rp (X1 = 20%)

Recession -0.07 0.17 0.122
Normal 0.12 0.07 0.08
Boom 0.28 -0.03 0.032

Given the R1 and R2 outcomes, we can use the portfolio result to compute the last
column. For example, in the recession case:

Rp = 0.2(−0.07) + 0.8(0.17) = 0.122.

For portfolios, just as for individual assets, we want to know my expected return and
risk. We answer these questions the same way we answered them last time:

• First, we can compute the mean directly (same way as we computed mean before):
1 1 1
R̄p = (0.122) + (0.08) + (0.032) = 0.078
3 3 3

• Also, recall that variance was the sum of squared deviations from the mean:

Scenario p R1 R1 − R̄1 R2 R2 − R̄2 Rp Rp − R̄p

Recession 1/3 -0.07 -0.18 0.17 0.10 0.122 0.044

Normal 1/3 0.12 0.01 0.07 0 0.08 0.002
Boom 1/3 0.28 0.17 -0.03 -0.10 0.032 -0.046

We can now find the variance of Rp :

1 1 1
σp2 = (0.044)2 + (0.002)2 + (−0.046)2 = 0.0014.
3 3 3

2
Jessica Wachter Class notes for Finance 604

• Recall variance is in %2 – it is easier to interpret its square root, standard

deviation: √
σp = 0.0014 = 0.037

• So, for our portfolio, we have learned that when X1 = 0.2 and X2 = 0.8:

R̄p = 0.078 and σp = 0.037.

Recall from last section:

R̄1 = 0.11, R̄2 = 0.07
and
σ1 = 0.143, σ2 = 0.082, Cov(R1 , R2 ) = −0.01167.

• The portfolio mean is 7.8%, which makes sense because it is between 11% and
7%. Also, it makes sense that it is closer to 7%, as we have 80% in the
corresponding asset.

• However, the standard deviation should seem a bit strange - it is below both the
standard deviation for R1 and R2 , which are 14.3%, and 8.2%, respectively.

We need to understand how the portfolio return can have a lower standard deviation
than both of its underlying assets. To do this, it is helpful to derive formulas that show
the mean and standard deviations of the portfolio in terms of R1 and R2 .

• Portfolio mean
R̄p = X1 R̄1 + X2 R̄2

• Portfolio variance
σp2 = X12 σ12 + X22 σ22 + 2X1 X2 σ1 σ2 ρ,
where ρ = correlation. To find the standard deviation, we take the square root.

The last formula can be written in terms of the covariance:

• Because we know:
Cov(R1 , R2 )
ρ= .
σ1 σ2
• Therefore:
Cov(R1 , R2 ) = σ1 σ2 ρ,

• so
σp2 = X12 σ12 + X22 σ22 + 2X1 X2 Cov(R1 , R2 ).

3
Jessica Wachter Class notes for Finance 604

These formulas will help us understand portfolios. They are also useful
computationally. While we can always compute the mean and standard deviation of Rp
from first principals as shown above, it is often faster to use the formulas.
The mean looks simple enough. In fact, it is identical to the returns formula. We
already knew something funny was going on with the variance from the last example, as
our portfolio has a standard deviation below that of its underlying components. Before
we prove these formulas, let’s recall our example:

• R̄1 = 0.11, R̄2 = 0.07.

• σ1 = 0.143, σ2 = 0.082.

• Cov(R1 , R2 ) = −0.01167.

Let X1 = 0.2 and X2 = 0.8. Therefore:

R̄p = 0.2(0.11) + 0.8(0.07) = 0.078.

σp2 = 0.22 (0.143)2 + 0.82 (0.082)2 + 2(0.2)(0.8)(−0.01177) = 0.0014.

So:
σp = 0.037.
This agrees with what we found before.
What does a negative correlation mean intuitively? Relative to their means, these two
assets’ returns move in opposite directions.
Thus, the formula suggests that it is very important for our portfolio that correlation is
negative. Look at the returns on the portfolio:

• In a recession, holding security 2 was very helpful – it offset the loss of security 1.

• In a boom, holding security 1 was very helpful – it offset the loss of security 2.

These shortcut formulas are pretty useful. Let’s see why they are true.
Proof of Formulas for the Mean and Variance of a Portfolio:
First, recall the rules of mean and covariance. For any random variables Z1 , Z2 , and Z3
and constant a:

1. E(Z1 + Z2 ) = E(Z1 ) + E(Z2 ).

2. E(aZ1 ) = aE(Z1 ).

4
Jessica Wachter Class notes for Finance 604

3. Cov(Z1 + Z2 , Z3 ) = Cov(Z1 , Z3 ) + Cov(Z2 , Z3 ).

4. Cov(aZ1 , Z2 ) = aCov(Z1 , Z2 ).

First, we prove the formula for the mean: from Rp = X1 R1 + X2 R2 , it follows from the
first rule above that:
E(Rp ) = E(X1 R1 ) + E(X2 R2 ).
Then, by the second rule:

E(Rp ) = X1 E(R1 ) + X2 E(R2 ).

This proves the formula for the mean.

To prove the formula for the variance, recall

σp2 = Cov(Rp , Rp ).

By repeated applications of the third rule:

Cov(Rp , Rp ) = Cov(X1 R1 + X2 R2 , X1 R1 + X2 R2 )
= Cov(X1 R1 , X1 R1 ) + Cov(X2 R2 , X2 R2 ) + 2Cov(X1 R1 , X2 R2 ).

Finally, applying the fourth rule:

Cov(Rp , Rp ) = X12 Cov(R1 , R1 ) + X22 Cov(R2 , R2 ) + 2X1 X2 Cov(R1 , R2 )

= X12 σ12 + X22 σ22 + 2X1 X2 Cov(R1 , R2 ).

This completes the proof of the variance formula.

Investment Opportunity Set
A typical problem regarding diversification between two risky assets would be as
follows: you are given securities R1 and R2 and know their means and standard
deviations. Now you want a menu of portfolio mean and standard deviation options for
the asset weights.

Example
R̄1 = 0.17 σ1 = 0.25
R̄2 = 0.10 σ2 = −.12
Correlation ρ = 0.2.
Security 1 has a higher expected return, but it is riskier than 2. Using the formulas, we
can compute the portfolio mean and standard deviation for a range of weights in R1
and R2 :

5
Jessica Wachter Class notes for Finance 604

X1 X2 R̄p σp
0 1 0.100 0.120
0.2 0.8 0.114 0.117
0.4 0.6 0.128 0.134
0.6 0.4 0.142 0.166
0.8 0.2 0.156 0.206
1 0 0.170 0.250

Consider a simple world where you run a mutual fund investing in these two assets. In
this case, this table gives a menu for your clients. For example, when X1 = 0.2 and
X2 = 0.8:
R̄p = 0.2(0.17) + 0.8(0.10) = 0.114
σp2 = (0.2)2 (0.25)2 + (0.8)2 (0.12)2 + 2(0.2)(0.8)(0.25)(0.12)(0.2) = 0.0136
Thus: √
σp = 0.0136 = 0.117

What’s striking about this? Even though security 2 is less risky than 1 when held by
itself, you can make your portfolio less risky by adding a bit of 1 ⇒ gain from
diversification.
Note: the correlation here is positive, not -1 as in the last example.
To make this point even more dramatic: an incredibly useful tool for thinking about
mean-variance analysis (that’s what this is called) is the mean-standard deviation
diagram - a graphical depiction of the menu you are laying out for your clients:

0.18

0.16
M
0.14

0.12

0.1
Rp

0.08
Rf
0.06

0.04

0.02

0
0 0.05 0.1 0.15 0.2 0.25
σp

6
Jessica Wachter Class notes for Finance 604

Every portfolio corresponds to a point on this diagram: this diagram has the mean of a
portfolio on its Y -axis and the standard deviation on its X-axis.
Diversification takes the form of this backward bend: you are getting a higher return,
but a lower standard deviation.
Note: You can derive this algebraically (we will not do this). R̄p is a linear function of
X1 . Thus, we can substitute in for X1 (note X2 = 1 − X1 ). σp2 is quadratic in the mean,
and σp is a hyperbolic function of the mean.
This backward bend means certain portfolios can be ruled out entirely.
Let’s take a step back and think about where your clients want to be (recall that they
prefer more wealth to less and are risk averse):

• Between points A and B, which is preferred?

Note: A and B have equal means, but B has a higher standard deviation. ⇒ A
dominates B.

• Between points A and C, which is preferred?

Note: A and C have equal standard deviations, but A has a higher mean. ⇒ A
dominates C.

7
Jessica Wachter Class notes for Finance 604

• What about B and C?

⇒ Neither dominates the other.

Implications for the Mean-Standard Deviation Diagram:

The entire lower portion of the bend can be disregarded - these points would never be
chosen by an investor who likes mean and dislikes standard deviation. In fact, points in
the upper portion are mean-variance efficient: for a given amount of risk (standard
deviation), they produce the highest expected return. Thus, these points dominate
those in the lower portion - you don’t need to even show the lower points to your client.
Which of the remaining points might your client pick? It depends on how risk averse
they are. The more risk averse, the lower down they go. This results in less risk, but
also in a lower mean. They pick the lowest variance if they are risk-obsessed.
Now we understand how we use this mean-standard deviation diagram. We would like
to also understand its shape - why does ρ = .2 produces a diversification benefit (now
that we understand why it is a benefit), and how it is that we can increase the mean
and decrease the standard deviation at the same time? To do so, we will look at three
special cases that are particularly intuitive: ρ = 1, ρ = −1 (total diversification is
possible), and ρ = 0.

8
Jessica Wachter Class notes for Finance 604

0.18
ρ=−1
0.17 ρ=0
ρ=1
0.16

0.15
E(Rp)
0.14

0.13

0.12

0.11

0.1
0 0.05 0.1 0.15 0.2 0.25
σp

1. Case I: Perfect Correlation ⇒ ρ = 1 (the solid line). Here, variance becomes:

σp2 = X12 σ12 + X22 σ22 + 2X1 X2 σ1 σ2 .

Note that correlation has gone away. Looking at the above equation, we see is
that this expression is in fact a perfect square:

σp2 = (X1 σ1 + X2 σ2 )2 .

This means:
σp = X1 σ1 + X2 σ2 .
⇒ In this special case, standard deviation is a weighted average, just like the
mean. We have two linear equations, so every point on our mean-standard
deviation diagram forms part of a straight line. As we move along the line, for
each bit of expected return we gain, we pay in terms of increased standard
deviation. Thus, in this case (and only with perfect correlation), there are no
gains from diversification.

2. Case II: Perfect Negative Correlation ⇒ ρ = −1 (the dotted line). This is

an entirely different story. We’ve already seen that some diversification s possible.
However, we didn’t realize how good it could be. In this case, variance becomes:

σp2 = X12 σ12 + X22 σ22 − 2X1 X2 σ1 σ2 .

Note this is also a perfect square:

σp2 = (X1 σ1 − X2 σ2 )2 .

9
Jessica Wachter Class notes for Finance 604

Because σp must be a positive number, this implies:

σp = |X1 σ1 − X2 σ2 |.

The amazing thing about ρ = −1 is that we can reduce the standard deviation all
the way to zero by cleverly choosing the portfolio weights:

X1 σ1 − X2 σ2 = 0.

X1 σ1 − (1 − X1 )σ2 = 0.
X1 (σ1 + σ2 ) − σ2 = 0.
σ2
X1 = .
σ1 + σ2
Therefore:
σ2 σ1 + σ2 − σ2 σ1
X2 = 1 − = = .
σ1 + σ2 σ1 + σ2 σ1 + σ2
Using the numbers from the problem:
0.12
X1 = = 0.32.
0.25 + 0.12
So:
X2 = 0.68.
This tells us that when ρ = −1, the plot must intersect the y-axis at 0.12. See the
dotted line in the diagram on page 11.
Thus, in the case of perfect negative correlation, we have huge gains in
diversification:

• This is what we said would happen with stocks and bonds in the three-state
example: when Asset 1 is high relative to its mean, 2 is low, and vice-versa.
• In this way, a portfolio that combines Asset 1 and 2 will not have as wide
signs as Asset 1 or Asset 2 by itself – the portfolio has less variance.
• What’s surprising: when two assets are perfectly negatively correlated, we
can reduce the risk of the combined portfolio all the way to 0.
• Note: by investing in asset 1 (stock), we still get the expected return benefit,
but we also get a big risk reduction.
• Note: we’ve also guaranteed ourselves a riskfree rate.

By the way, what is the mean of this portfolio?

R̄p = 0.32(0.17) + 0.68(0.10) = 0.12

10
Jessica Wachter Class notes for Finance 604

3. Case III: Independence ⇒ ρ = 0. This last case is crucial to our

understanding of gains from diversification. It may at first seem like we don’t
have gains. However, we do in fact have gains from diversification here too.
Consider what happens to variance when we put in ρ = 0:

σp2 = X12 σ12 + X22 σ22 .

This looks like a weighted average, but these are in fact squares, and they sum to
less than one. In fact:

σp = (X12 σ12 + X22 σ22 )1/2 < X1 σ1 + X2 σ2 .

We see that this standard deviation is much less than the case of ρ = 1. The best
way to prove it to yourself is to plug in the numbers. Let’s do:

X1 X2 R̄p σp (ρ = 1) σp (ρ = 0)
0 1 0.100 0.120 0.120
0.2 0.8 0.114 0.146 0.108
0.4 0.6 0.128 0.172 0.123
0.6 0.4 0.142 0.198 0.158
0.8 0.2 0.156 0.224 0.201
1 0 0.170 0.250 0.250

To see how this diagram would play out, see the dashed line on the graph on page
11.
While this is not as dramatic as ρ = −1, we can simultaneously increase our mean
and decrease our standard deviation ⇒ we still get huge gains from diversification.
Now suppose we put 20% in security 1 and 80% in 2. Then, when ρ = 1,
σp = 0.146 (which is between the two and closer to 1). When ρ = 0, σp = 0.108.
This standard deviation is much smaller. Note that the mean in the same in both
cases (0.114).
There are gains from diversification here too. Where are they coming from?
To understand, this we need to go one level back and revisit a remarkable
statistical fact about independent coin flips. Consider two independent coins,
where you earn:

• 100% if coin 1= heads, -50% if tails

• 100% if coin 2=heads, -50% if tails

You could bet everything on one coin:

1/2 h 100%
1/2 t -0%

11
Jessica Wachter Class notes for Finance 604

Or you could divide your wealth among the two coins:

1/4 (h,h) 0.5(1) + 0.5(1) 100%
1/4 (h,t) 0.5(1) + 0.5(0) 50%
1/4 (t,h) 0.5(0) + 0.5(1) 50%
1/4 (t,t) 0.5(0) + 0.5(0) 0%
In both cases, the mean is:
1 1
(1) + (0) = 50%
2 2
⇒ It all comes down to what your grandma told you. Don’t put all your eggs in
one basket. That way, if one basket breaks, you still have the other eggs.

4. CASE IV: Real Life ⇒ ρ > 0 In real life, the most common case we encounter
is ρ > 0. For example, in the case above, ρ = 0.2. The intuition for the positive
correlation case is exactly the same as for zero correlation, though the gains from
diversification are less powerful. The graph below depicts both ρ = 0.2 and
ρ = 0.5:

0.18
ρ=−1
0.17 ρ=0
ρ=.2
ρ=.5
0.16
ρ=1

0.15
E(Rp)

0.14

0.13

0.12

0.11

0.1
0 0.05 0.1 0.15 0.2 0.25
σp

Note that for ρ = .5, there is no backward bend, and standard deviation steadily
increases. However, there are still gains from diversification, relative to the case of
perfect correlation. The standard deviation of the portfolio is still less than a
weighted average of the standard deviations of the component parts.

• How large can portfolio risk become, at any level of correlation?

Suppose I gave you the set-up that we started with:

R̄1 = 0.17 σ1 = 0.25

R̄2 = 0.10 σ2 = 0.12

12
Jessica Wachter Class notes for Finance 604

However, I didn’t tell you the correlation. Instead, I told you that I planned to
put:

X1 = 0.6
X2 = 0.4

(Note: the point below holds for any weights, not just 60%/40%.)
How large can my portfolio standard deviation be? Is there an upper bound, or
can it get arbitrarily high? Let’s calculate standard deviation for different values
of ρ:
ρ = −1 ρ = 0 ρ = 0.2 ρ = 0.5 ρ = 1
R̄p 0.142
σp 0.102 0.158 0.166 0.169 0.198
First, notice that means stay the same, but standard deviations increase. Why?
Because correlation doesn’t enter into the formula for the mean. Second, notice
that gains from diversification decrease as we get to ρ = 1.
We can plot these points on the diagram (this is the same diagram as in Case IV):

0.18
ρ=−1
0.17 ρ=0
ρ=.2
ρ=.5
0.16
ρ=1

0.15
E(Rp)

0.14

0.13

0.12

0.11

0.1
0 0.05 0.1 0.15 0.2 0.25
σp

From this graph, we can see:

– The higher the standard deviation, the higher ρ becomes:
1 1
σp = (X12 σ12 + X22 σ22 + 2X1 X2 σ1 σ2 ρ) 2 ≤ (X12 σ12 + X22 σ22 + 2X1 X2 σ1 σ2 ) 2 .

– When ρ = 1, standard deviation is a weighted average:

1
(X12 σ12 + X22 σ22 + 2X1 X2 σ1 σ2 ) 2 = X1 σ1 + X2 σ2 .

– Putting these facts together:

σ2 ≤ X1 σ1 + X2 σ2 = 0.6(0.25) + 0.4(0.12) = 0.198.

13
Jessica Wachter Class notes for Finance 604

Thus, diversification can never create a “monster portfolio.” That is, standard
deviation never “explodes” because it can never exceed a weighted average of the
underlying portfolio assets’ standard deviations, as ρ can never be greater than
one.
Thus, the largest possible standard deviation is a weighted average of σ1 and σ2 .

(b) Portfolios with One Riskless and One Risky Asset

So far, we have assumed two risky assets and looked at the shape of the investment
opportunity set for various correlations. What happens if one of the assets is riskless
(riskfree)? By that, I mean that we know for certain exactly what it will return next
year.
Recall from our holding period return discussion that the riskless asset is the
zero-coupon bond with maturity equal to the investor’s holding period. Let’s say we
have a holding period of one year. Then, the riskless asset is a one-year Treasury Bill.

Example Assume you hold two assets: a Treasury Bill and an S&P 500 mutual fund :

1. S&P Mutual Fund: R̄M = 15%, σM = 22%.

2. T-bill: Rf = 7%, σf = 0.

We denote the return on the riskfree asset as Rf . We will show what the mean-standard
deviation diagram looks like in this case. It turns out to be a straight line:

0.2
M

0.15
Rp

0.1

Rf

0.05

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
σp

Why is this? Let’s see what assuming a riskless asset does to our formulas:

14
Jessica Wachter Class notes for Finance 604

1. Mean:
R̄p = X1 R̄1 + X2 R̄2 .
In this case:
R̄p = Xf R̄f + XM R̄M .
Note: because the T-Bill return is riskfree, the actual return is equal to the mean:

R̄p = Xf Rf + XM R̄M .

2. Standard deviation: also, we know that:

X f = 1 − XM .

Substituting in:
R̄p = (1 − XM )Rf + XM R̄M .
The formula for variance becomes:

σp2 = Xf2 σf2 + XM

2 2
σM + 2ρXM Xf σM σf

However, σf = 0. Therefore:1

σp2 = XM
2 2
σM .
σp = XM σM .

This formula works as long as XM ≥ 0, which we will assume.

Using these results, we can show that the portfolio set forms a straight line on the
mean-standard deviation diagram.

• Rearranging the second equation above tells us:

σp
XM = .
σM

• Thus:  
σp σp
R̄p = 1− Rf + R̄M .
σM σM

• Finally:
R̄M − Rf
R̄p = Rf + σp .
σM
1
We can also come to this conclusion by noting that Cov(Rf , RM ) = 0 because Rf is a constant.

15
Jessica Wachter Class notes for Finance 604

This describes a straight line relating R̄p to σp with:

• Intercept: Rf .
• Slope:
R̄M − Rf
.
σM
This slope is also frequently called the Sharpe ratio. (Bill Sharpe won the Nobel
prize for the CAPM, which we will soon see). This ratio represents the reward
(expected return) you earn for each unit of risk (standard deviation) you take on.

This line describes the entire investment universe with one risky and one riskless asset.
More risk averse folks hold relatively more of the risky asset. Who will hold the
portfolio made entirely of Rf ? The risk-obsessed person. Let’s look at a few portfolio
examples.

Example Assume: RM = 0.15, Rf = 0.07, and σM = 0.22. Also, XM = 0.8.

• Therefore, Xf = .2.
• What’s the mean of your portfolio?

R̄p = (1 − XM )Rf + XM R̄M

= 0.2(0.07) + 0.8(0.15)
= 0.134.

• What’s the standard deviation?

σp = XM σM
σp = 0.8(0.22)
= 0.176.

Example Assume: we want to have σp = 0.11.

• What fraction do I invest in the market?

σp 0.11
XM = = = 1/2.
σM 0.22

• What fraction does this imply in the riskfree asset?

Xf = 1 − XM = 1/2.

16
Jessica Wachter Class notes for Finance 604

• What’s the expected return on the portfolio?

R̄p = 1/2(0.15) + 1/2(0.07) = 0.11

(The 0.11 here is a coincidence.)

• How else could I have figured out the mean?

R̄M − Rf
R̄p = Rf + σp
σM
= 0.07 + (0.11)0.37
= 0.11

Note that we’ve drawn the line going through RM . This point represents everything in
the riskfree asset, RM . Algebraically, the points north of RM represent XM > 1,
Xf < 0. What do these points represent economically? Note that:

• A very risk averse person: will choose a portfolio consisting mostly of the riskfree
asset.

• A less risk averse person: will be more towards putting all her wealth in the
mutual fund.

• What about the person who is even less risk averse? Suppose the mutual fund is
too tame for you. More generally, can we achieve a higher mean than R̄M ? How?

– This corresponds to borrowing at Rf and investing the proceeds in the

mutual fund.
– Another terms for this include buying shares on margin and using leverage.

Let’s look at an example understand why our portfolio formulas still hold.

Example Suppose we set XM = 1.5. Then, Xf must be: Xf = −0.5. We will go

through the transaction set that implies these numbers and prove from first principles
that the return is what the formulas say it should be.
Suppose your wealth is $1000 ⇒ borrow $500 at Rf and invest $1500 at RM :

• Next year, on your mutual fund investment, you receive:

$1500(1 + RM ).

17
Jessica Wachter Class notes for Finance 604

• However, you have to pay back what you borrowed with interest. So, you owe:

$500(1 + Rf ).

• Therefore, your expected holding period return is:

$1500(1 + RM ) − $500(1 + Rf )
R̄p = −1
$1000
= 1.5 + 1.5RM + .5 + (−0.5)Rf − 1
= 1.5RM + (−0.5)Rf .

Note: the second equation follows from dividing through by $1000.

Now that we understand the economics, we can use the formulas in exactly the same
way.
Let’s go back to the example:

R̄p = (1.5)(0.15) + (−0.5)(0.07) = 0.19

σp = (1.5)(0.22) = 0.33
⇒ Not a surprise that both the mean and standard deviation are higher.

(c) Optimal Portfolios with One Riskless and Two Risky

Assets

Before we combine everything, assume a choice of risky assets to combine with the
riskfree:

• Risky mutual fund A: low expected return, low standard deviation.

• Risky mutual fund B: higher expected return, higher standard deviation.

We can invest in A or B in combination with Rf . Which would we rather combine with

Rf ? Let’s look at the investment opportunity set in each case: Note: portfolios
combining Rf with B lie on the solid line, and those combining Rf with A lie on the
dashed line.

18
Jessica Wachter Class notes for Finance 604

0.2

0.18

0.16 B
0.14

0.12
Rp
0.1
A
0.08
Rf
0.06

0.04

0.02

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
σp

The diagram makes it clear that you would rather combine Rf with B, as the resulting
portfolios have a lower standard deviation for a given mean than the portfolios of A
with Rf . ⇒ Regardless of your risk preferences, you would rather combine Rf with B.
Mathematically, why is B so good? The line between B and Rf has a higher slope than
that between A and Rf . In other words: B has a higher Sharpe ratio .
Generally, it is optimal to combine Rf with the risky asset with the highest slope.

Example
R̄A = 0.10 σA = 0.12
R̄B = 0.17 σB = 0.25
Suppose you can’t form a portfolio. Which would you rather combine with Rf ?
You could plot the lines. However, it is even easier to just calculate their slopes:

• Sharpe ratio for A:

0.10 − 0.07
= 0.25
0.12
• Sharpe ratio for B:
0.17 − 0.07
= 0.40
0.25

Note: the Sharpe ratio for B is higher. However, we cannot tell which is better simply
by looking at the mean and variance. We have to compute the slope.

It is not hard to come up with an example where the asset with lower mean and lower
standard deviation has a higher Sharpe ratio:

19
Jessica Wachter Class notes for Finance 604

Note how important leverage is here to making C attractive to less risk averse investors.
Let’s combine the two cases we’ve studied so far: this is most realistic. You have access
to a large universe of risky assets. Also, you can combine RB and RA . Consider the
minimum variance frontier for the risky assets (made up of portfolios of risky assets
yielding the lowest variance for a given mean). You can combine Rf with any portfolio
on the frontier:

0.18

0.16
M
0.14

0.12

0.1
Rp

0.08
Rf
0.06

0.04

0.02

0
0 0.05 0.1 0.15 0.2 0.25
σp

Which is the best portfolio to combine with Rf ?

• Clearly, this would be the tangency portfolio because it has the highest slope.

• Any portfolios with a lower slope will be sub-optimal and inefficient.

• Any portfolios with higher slope cannot be attainable.

What has the addition of the Rf done to the set of portfolios investors want to hold?

20
Jessica Wachter Class notes for Finance 604

• Recall investment opportunities with only two risky assets. Then, portfolios in
the lower portion of this curve were dominated, or inefficient. Thus, portfolios in
the upper portion were mean-variance efficient.

• Now that we’ve introduced Rf , the entire curve except for one point is inefficient.

• The new efficient set is equal to the line between the riskfree rate and the
tangency portfolio. The points along this line now dominate the previously
efficient portfolios.

⇒ The line between Rf and the tangency portfolio is the capital allocation line. Any
investor who likes mean and dislikes variance will want a portfolio along this line.
Let’s now look at an example of the implications of the capital allocation line:

Example Suppose you’ve already found the tangency portfolio. It consists of:

Asset Weight
Asset A 0.6
Asset B 0.4
Riskfree asset 0

XA = 60%, XB = 40%.
Remember there is a riskfree asset. If you hold the tangency portfolio, your weights are:
R1 = 0.6, R2 = 0.4, and Rf = 0.
Suppose, however, the tangency portfolio is a bit too risky for you. You would rather
combine it with Rf . You hold 70% in the tangency portfolio. What are your weights?

XA = 0.7(0.6) = 0.42.
XB = 0.7(0.4) = 0.28.
XRf = 0.3.

⇒ Because everyone holds some combination of Rf and the tangency portfolio,

everyone will hold the risky assets in the same proportion:

0.6/0.4 = 3/2 and 0.42/0.28 = 3/2.

21
Jessica Wachter Class notes for Finance 604

(d) Optimal Portfolio: One Riskless and Multiple Risky

Assets

How do our proofs change when we go from 2 to N assets, where N is some large
number?
We wil return, for a moment, to the case without a riskless asset. Suppose we’ve got a
third portfolio, fund C. This fund has high variance and low mean ⇒ this makes it look
very unattractive when held by itself.
In fact, B dominates it. Nobody wants to hold C on its own. However, suppose C also
has a low correlation with B. Thus, you would want to hold it as part of a portfolio.

Observe how portfolio C moves the curve representing the investment opportunity set
further to the left.
Our discussion above covers three assets. In general, we have lots of assets, and we can
spend all day drawing frontiers that connect them.
Instead of doing this, we simplify matters by asking ourselves, what do we really want
to know? We want to find, for a given mean return, the portfolio with the lowest
possible standard deviation. This leads us to the following definition:

Definition A portfolio is minimum variance if it minimizes the variance for a given

mean.

We can show the set of minimum variance portfolios (called the minimum variance
frontier, because its the furthest out you can go) on a mean-standard deviation diagram.
Finding minimum variance portfolios for an arbitrary number of assets requires
multivariate calculus and linear algebra. We will not go into it here. However, we can
show that the shape of the frontier must be concave.
Suppose instead of being concave that the shape looked like:

22
Jessica Wachter Class notes for Finance 604

Then we could form a portfolio by considering the portfolios at the inflection points.
It is also useful to define the point that minimizes the variance out of the entire
universe of risky asset portfolios:

Definition The Global Minimum Variance Portfolio (GMVP) has the lowest possible
variance of any portfolio.

Note that any portfolios that lie in the bottom half of the curve are dominated by those
in the top half. Moreover, all portfolios are dominated by the top half. We will call the
top half (namely, those portfolios with mean returns larger than that of the GMVP) the
efficient frontier.
As in our example with three assets, an individual asset, even if it is dominated by
other assets, may still be part of an efficient portfolio.
Now that we know that the frontier is concave, we also know that adding Rf works the
same way as before. Thus, what is the best portfolio in this entire portfolio universe to
combine with Rf ? As before, it is the tangency portfolio, the one with the highest
Sharpe ratio.
We will call the line that connects Rf with the tangency portfolio the Capital
Allocation Line.
This leads us to the Mutual Fund Theorem! Note that, like the theorem from the
Introduction, it is a separation theorem.

Theorem (Mutual Fund Theorem). The portfolio allocation problem can be divided
into two steps.

1. Determine the tangency portfolio ⇒ this is the optimal combination of the risky
assets, or the Mutual Fund.

2. Determine where you want to be on the capital allocation line.

Why is this theorem so amazing?

23
Jessica Wachter Class notes for Finance 604

• Any two investors who agree on this diagram (namely, agree on the means,
variances, and correlations) will locate their portfolios along the capital allocation
line. In other words, these investors should hold all risky assets in the same
proportions.

• They will choose to invest in some combination of the tangency portfolio and Rf .

• An investment adviser’s job becomes very easy. Everyone’s tangency portfolio is

the same, so you just need to find it, then determine where your clients fall along
this line.

• This is called the mutual fund theorem because it predicts that as long as
investors have the same beliefs, they will all invest in an identical mutual fund, in
combination with a riskfree security.

In general, what characteristics will lead an asset to have a high weight in the tangency
portfolio?

1. High expected return.

2. Low standard deviation.

3. Low correlation with the other securities.

24