FE 445 – Investment Analysis and Portfolio
Management
Fall 2020
Farzad Saidi
Boston University | Questrom School of Business
Lecture 6: Diversification
Return of a portfolio
n
X
rp = wi ri ,
i=1
where
Pn
• wi is the proportion of funds in security i, where: i=1 wi = 1
• ri is the return on security i
• n is the number of securities
Example: You invested V0 = $10, 000 at the beginning of the year. You
have put
• $7,000 into asset A: the annual return was 10%
• $3,000 into asset B: the annual return was 20%
There are two ways of calculating the portfolio returns:
1. V1 = $7, 000 × 1.1 + $3, 000 × 1.2 = $11, 300. Hence rp = 13%
2. rp = wA rA + wB rB = 0.7 × 10% + 0.3 × 20% = 13%
1
Diversification with many assets
• n is the number of securities in the portfolio
• portfolio weights are w = 1/n for each asset
2
Two-security portfolio return
We have shown that
rp = wA rA + wB rB .
Since this holds for every realization (state), this implies that
E (rp ) = wA × E (rA ) + wB × E (rB )
and
Var(rp ) = σ 2 (rp ) = wA2 × σA2 + wB2 × σB2 + 2wA wB Cov(rA , rB )
= wA2 × σA2 + wB2 × σB2 + 2wA wB Corr(rA , rB )σA σB .
3
Covariance
Definition:
S
X
Cov(rA , rB ) = p(s) × (rA (s) − E (rA )) × (rB (s) − E (rB )) ,
s=1
where s are the different scenarios that happen with probability p(s)
• This almost looks like the formula for variance. Recall that
S
X 2
Var(r ) = σ 2 = p(s) × (r (s) − E (r ))
s=1
Except now you multiply the return deviations of two returns instead
of squaring them
⇒ Covariance
• It measures the comovement of two returns
4
Ex-post covariance calculation
When using historical data, then
n
1 X
Cov(r1 , r2 ) = (r1,t − r¯1 ) × (r2,t − r¯2 ) ,
n−1
i=1
where
• r¯1 is the average return for asset 1
• r¯2 is the average return for asset 2
• n is the number of observations
Evaluating the covariance term by term:
r1 − r¯1 r2 − r¯2 (r1 − r¯1 ) × (r2 − r¯2 )
+ + +
– – +
+ – –
– + – 5
Example
Economy has two states:
1. Good with probability pg = 0.7
2. Bad with probability pb = 0.3
state prob rA rA − E (rA ) rB rB − E (rB ) rC rC − E (rC )
good 0.7 15% 5% -12%
bad 0.3 -5% -5% 15%
• Cov (rA , rB ) = . . . . . . . . .
• Cov (rA , rC ) = . . . . . . . . .
6
Correlation
• Problem with covariance: covariance depends on standard deviation
of returns, not only on comovement
• Correlation coefficient: standardize covariance by standard
deviations to get degree to which the stocks move together
Cov(rA , rB )
ρA,B = Corr(rA , rB ) =
σA σB
• By definition: −1 ≤ ρA,B ≤ +1 and ρA,B = ρB,A
• ρA,B = 1: returns perfectly move together
• ρA,B = −1: returns move exactly opposite
• ρA,B = 0: returns are unrelated (independent): dice roll
7
Examples: Apple and JP Morgan
0.15
0.1
0.05
returns
-0.05 APPL
JPM
-0.1
Jan17 Apr17 Jul17 Oct17 Jan18
date
Is the correlation high?
8
Examples: Tesla and JP Morgan
0.15
0.1
0.05
return
-0.05
-0.1 TESLA
JPM
-0.15
Jan17 Apr17 Jul17 Oct17 Jan18
date
Is the correlation high?
9
Examples: bank stocks
0.15
0.1
0.05
returns
0
BoA
-0.05 GS
JPM
-0.1
Jan17 Apr17 Jul17 Oct17 Jan18
date
Is the correlation high?
10
In Excel
Bank of America Goldman Sachs
date price return price return
1/1/2017 22.64 226.47
2/1/2017 24.68 0.09 244.97 0.08
3/1/2017 23.59 -0.05 227.46 -0.07
4/1/2017 23.34 -0.01 221.59 -0.03
5/1/2017 22.41 -0.04 209.18 -0.06
6/1/2017 24.26 0.08 220.45 0.05
7/1/2017 24.12 -0.01 223.86 0.02
8/1/2017 23.89 -0.01 222.28 -0.01
9/1/2017 25.34 0.06 236.45 0.06
10/1/2017 27.39 0.08 241.72 0.02
11/1/2017 28.17 0.03 246.86 0.02
12/1/2017 29.52 0.05 254.76 0.03
1/1/2018 32.20 0.09 268.14 0.05
=COVARIANCE.S(C4:C15,E4:E15) cov 0.0022
=CORREL(C4:C15,E4:E15) corr 0.9153
• Excel divides with 1/n instead of 1/(n − 1) in COVAR
• Data from finance.yahoo.com: click on “historical prices” then
“download to spreadsheet”, save as .csvfile
• Use adjusted close price to calculate returns
11
Examples: bank stocks
0.15
0.1
0.05
returns
0
BoA
-0.05 GS
JPM
-0.1
Jan17 Apr17 Jul17 Oct17 Jan18
date
Portfolio = 31 JPM + 13 BOA + 31 GS ⇒ Can you diversify?
12
Diversification with two risky assets?
Recall that the variance of portfolio formed using two assets is:
σp2 = wA2 σA2 + wB2 σB2 + 2wA wB Cov(rA , rB ),
and since Cov(rA , rB ) = σA σB ρA,B :
σp2 = wA2 σA2 + wB2 σB2 + 2wA wB σA σB ρA,B .
If σB = 0 (because it is the risk-free asset), the formula simplifies to
σp = w A σA .
There are diversification benefits if:
wA σA + wB σB > σp
Why?
• If ρA,B = 1, there are no diversification benefits
⇒ σp2 = (wA σA + wB σB )2
• If ρA,B < 1, some diversification benefits
⇒ Lower correlation implies more diversification! 13
The power of naı̈ve diversification
Naı̈ve ≡ just putting randomly chosen stocks together in an equally
weighted portfolio
Most of the diversifiable risk eliminated after 20 − 25 or so stocks!
14
Example: two-security portfolio
Two stocks:
σA2 = 0.15265, σB2 = 0.17543, Cov(rA , rB ) = 0.05933
• Let wA = 60% and wB = 40%. What is the volatility of the
portfolio?
• Did you get diversification benefits?
15
Creating a risk-free asset through perfect diversification
Scenario: Stock A offers 8% return at 40% std. dev., stock B offers 13%
return at 60% std. dev., the two are perfectly negatively correlated
(ρA,B = −1)
⇒ What is the risk-free rate in equilibrium?
16
Creating a risk-free asset through perfect diversification
Scenario: Stock A offers 8% return at 40% std. dev., stock B offers 13%
return at 60% std. dev., the two are perfectly negatively correlated
(ρA,B = −1)
⇒ What is the risk-free rate in equilibrium?
• As ρA,B = −1,
σp2 = wA2 σA2 + wB2 σB2 − 2wA wB σA σB = (wA σA − wB σB )2
• Risk-free asset = zero variance, so:
0 = 0.4wA − 0.6(1–wA ) ⇒ wA = 0.6
• The risk-free rate is thus equal to: 0.6 × 0.08 + 0.4 × 0.13 = 10%
16
Summary
This class:
• Covariance
• Diversification
Next class:
• Optimal diversification
• Efficient mean-variance frontier
17
