CHAPTER 6

Efficient Diversification
Objectives:
1/. How covariance & correlation affect power of
diversification to reduce portfolio risk.
2/. Construct efficient portfolios.
3/. Calculate composition of the optimal risky
portfolio.
4/. Use index model to analyze risk characteristics of
securities & portfolios.
McGraw-Hill/Irwin © 2008 The McGraw-Hill Companies, Inc., All Rights Reserved.
6.1 DIVERSIFICATION AND
PORTFOLIO RISK

6-2
Diversification and Portfolio Risk

Market risk
– Systematic or Non-diversifiable
– i.e.: macroeconomics (business cycle,
inflation, interest rate, exchange rate,…
Firm-specific risk
– Diversifiable or nonsystematic
– i.e.: industry influencing factors, company’s
R&D result, management style & philosophy,
…

6-3
 Figure 6.1 Portfolio Risk as a
Function of the Number of Stocks

6-4
 Figure 6.2 Portfolio Risk as a
Function of Number of Securities

Calculation:
Average SDs of
equally weighted
portfolios
constructed by
selecting number
of stocks in
random

6-5
6.2 ASSET ALLOCATION WITH
TWO RISKY ASSETS

6-6
 Covariance and Correlation
1. Covariance: a measure of the extent to which the
returns tend to vary with each other, that is, to co-
vary.
2. Correlation coefficient: measure the power of
relationship between couple of returns on two
portfolios

6-7
 Two Asset Portfolio:
Return – Stock and Bond

rp  wrB B
 wS r S
rP  Portfolio Return
wB  Bond Weight
rB  Bond Return
wS  Stock Weig ht
rS  Stock Return

6-8
 Two Asset Portfolio:
Covariance and Correlation Coefficient
Covariance:
S
Cov(rS , rB )   p (i ) rS (i )  rS  rB (i )  rB 
i 1

Correlation Coefficient:

Cov(rS , rB )
 SB 
 S B
6-9
Correlation Coefficients: Possible Values
Range of values for  1,2 : -1.0 < < 1.0
1. Portfolio risk (variance) is lower when  is lower.
2. If = 1.0, the securities would be perfectly positively
correlated.
3. If = - 1.0, the securities would be perfectly negatively
correlated
4. If = 0, the securities would be unrelated to each other.
Implications:
If correlation b/t components securities is small or negative,
then there will be a greater tendency for the variability in
the returns on the two assets to offset each other. Hence,
reduce portfolio risk.
6-10
Two Asset Portfolio (Stock and Bond):
Standard Deviation

  w   w   2w w   
2 2 2 2 2
p B B S S B S S B B,S

  Portfolio Variance
2
p

  Portfolio Standard Deviation

2
p

6-11
Three Rules of Two-Risky-Asset Portfolios
1. Rate of return on the portfolio:

rP  wB rB  wS rS
2. Expected rate of return on the portfolio:

E (rP )  wB E (rB )  wS E (rS )

3. Variance of the rate of return on the portfolio:

 P2  ( wB B ) 2  ( wS S ) 2  2( wB B )( wS S )  BS

6-12
In General, For an n-Security Portfolio:
rp = Weighted average of the
n securities
p2 = (Consider SUM of all
weighted pair-wise
covariance measures)

6-13
 I.e.: performance of a portfolio
invested in the stock & bond funds.

Spreadsheet 6.1 / 6.2 / 6.3 / 6.4 in page

149 & 150.

6-14
 Figure 6.4 Investment Opportunity Set for
Stocks and Bonds with Various Correlations
1. P = +1:
σ p = w b σb + w s σ s
Diversification is not effective
σp = 0 when short sell stock:
- ws = σb / (σp - σb ) & wb = 1+ ws

2. P < 1:
σ2p = (wb σp + ws σs)2 - ∆wbσpwsσs
σ p < w b σp + w s σ s
Diversification makes portfolio SD is
less than weighted average of
SD of component securities.

3. P = -1
σp = ABS[wb σb - ws σs]
Benefits from diversification stretch
to the limits.
σp = 0 when: wb = σs / (σb + σs )
& wb = 1- ws 6-15
 Using historical Data to estimate
expected portfolio risk & return
1. Approach 1: scenario analysis.
2. Approach 2: use of historical data.
– Useful prediction for near-future – perhaps next month
or next quarter in assumption that: variability & co-
variability change slowly over time.
– Expected indicators of simple security:
1. Mean = average value of sample historical period
2. Variance = average value of squared deviations around the
sample mean.
3. Covariance = average value of cross-product of deviation

6-16
 Example 6.1 (page 152)
Rate of return Deviation from Average return Product Example 6.1
Year Stock Fund Bond Fund Stock Fund Bond Fund of Deviations Using historical data
2006 30.17 5.08 20.17 0.08 1.53 to estimate means,

2007 32.97 7.52 22.97 2.52 57.78 Standard Deviations,

2008 21.04 -8.82 11.04 -13.82 -152.56 Covariance & Correlation

2009 -8.1 5.27 -18.10 0.27 -4.82

Conclusion:
2010 -12.89 12.2 -22.89 7.20 -164.75
1. Variance &
2011 -28.53 -7.79 -38.53 -12.79 493.00 covariance
2012 22.49 6.38 12.49 1.38 17.18
estimates: reliable
for short term.
2013 12.58 12.4 2.58 7.40 19.05
2. Future expected
2014 14.81 17.29 4.81 12.29 59.05 returns: highly
2015 15.5 0.51 5.50 -4.49 -24.70
noisy
3. In practice,
macroeconomic &
Average 10.00 5.00 Covariance 30.08 industry analysis
SD 19.00 8.00 Correlation 0.20
improves
estimates of
mean return.
Result of spreadsheet 6.1: (1) as the same for expected of average return rate & its SD

(2) Correlation quite different (scenario analysis: -0.49)

6-17
 Risk-return trade-off with two-risky-
assets portfolios
Input Data Spreadsheet 6.5
E(rS) E(rB) S B BS The investment opportunity set
10 5 19 8 0.2 with the stock and bond funds
Portfolio Weights Expected Return, E(rP) Std Dev
wS = 1–wB wB Col A*A3 + Col B*B3 Equation 6.6
-0.2 1.2 4.0 9.59 Conclusion:
-0.1 1.1 4.5 8.62
Beginning w/t 100% bond
0.0 1.0 5.0 8.00
0.0932 0.9068 5.5 7.804
portfolio: taking more
0.1 0.9 5.5 7.81 volatile assets (stocks)
0.2 0.8 6.0 8.07 actually reduces portfolio
0.3 0.7 6.5 8.75 risk (SD), while rise up
0.4 0.6 7.0 9.77 expected return, until the
0.5 0.5 7.5 11.02
minimum-variance set of
0.6 0.4 8.0 12.44
0.7 0.3 8.5 13.98
investment opportunity
0.8 0.2 9.0 15.60 reaches.
0.9 0.1 9.5 17.28
1.0 0.0 10.0 19.00
1.1 -0.1 10.5 20.75
1.2 -0.2 11.0 22.53
Notes:        
1. Negative weights indicate short positions  
2. The weights of the minimum-variance portfolio are  
computed using the formula in footnote 1 (page 154)  
6-18
Figure 6.3 Investment Opportunity
Set for Stocks and Bonds
1. Portfolios lie below
the minimum-
variance portfolio in
the figure is rejected
as inefficient.

2. Best choice among

portfolios on the
upward-sloping
portion of the curve is
not obvious, because
in this region higher
expected return is
accompanied by
greater risk.
6-19
 Spreadsheet 6.6 (page 157)
Input Data
E(r S ) E(r B ) sS sB
10 5 19 8

1
Weight in stocks Portfolio expected return Portfolio Standard Deviation for Given Correlation, r
wS E(rP) = Col A*A3 + (1 - Col A)*B3 -1 0 0.2 0.5 1
-0.1 4.5 10.70 9.00 8.62 8.02 6.90
0.0 5.0 8.00 8.00 8.00 8.00 8.00
0.1 5.5 5.30 7.45 7.81 8.31 9.10
0.2 6.0 2.60 7.44 8.07 8.93 10.20
0.3 6.5 0.10 7.99 8.75 9.79 11.30
0.4 7.0 2.80 8.99 9.77 10.83 12.40
0.6 8.0 8.20 11.84 12.44 13.29 14.60
0.8 9.0 13.60 15.28 15.60 16.06 16.80
1.0 10.0 19.00 19.00 19.00 19.00 19.00
1.1 10.5 21.70 20.92 20.75 20.51 20.10
2,3,4,5
Minimum Variance Portfolio
wS(min) = (sB ^2 - sBsSr ) / (sS^2 + sB^2 - 2* sBsSr) = 0.2963 0.1506 0.0923 -0.0440 -0.7273
E(rP) = wS(min)*A3+(1-w S (min))*B3 = 6.48 5.75 5.46 4.78 1.36
sP = 0.00 7.37 7.80 7.97 0.00
Notes:
1. s P = SQRT[ (Col A*C3)^2 + ((1 - Col A)*D3)^2 + 2*Col A*C3*(1 - Col A)*D3* r ]
2. The standard deviation is calculated from equation 6.6 using the weights of the miniumum-variance portfolio:
3. As the correlation coefficient grows, the minimum variance portfolio requires a smaller position in stocks (even a
negative position for higher correlations), and its performance becomes less attractive.
4. Notice that with correlation of .5 or higher, minimum variance is achieved with a short position in stocks.
The standard deviation is lower than that of bonds, but the mean is lower as well.
5. With perfect positive correlation (column G), you can drive the standard deviation to zero by taking a large,
short position in stocks. The mean return is then as low as 1.36%.
6-20
Methodology to combine
risk-free assets with
risky assets portfolio in
term of investment
efficiency
Methodology to combine risk-free assets with

Two-risky-assets portfolio construction

1. Identify the best possible or most efficient-risk-
return combinations available from the universe of
risky assets.
2. Determine the optimal portfolio of risky assets –
steepest CALO (maximizes its Sharpe ratio w/t Rf).
3. Choose an appropriate complete portion on
optimal CALO based on the investor’s risk aversion
by mixing the risk-free assets w/t optimal portfolio
(O).

6-22
6.3 THE OPTIMAL RISKY PORTFOLIO
WITH A RISK-FREE ASSET

6-23
Extending to Include Riskless Asset

The optimal combination becomes linear

A single combination of risky and riskless
assets will dominate.
Capital allocation line (CAL): the straight line
present the combination of one risk-free asset
with a risky portfolio of bonds & stocks.
Slop of the CAL = Sharpe ratio (reward-to-risk)
= [E(rp) – rf] / σp

6-24
 Figure 6.5 Opportunity Set Using Stocks
and Bonds and Two Capital Allocation Lines

6-25
 Dominant CAL with a Risk-Free
Investment (F)
1. CAL(O) reaches ultimate point of tangency with the investment
portfolio of bonds & stocks – it dominates other lines - it has the
largest slope (highest reward-to-volatility ratio)

Slope CAL(A) = reaches highest CAL(O)

E (rA )  rf
2. Composition of the optimal risky portfolio O that maximize the
A
portfolio’s Sharpe ratio:
Wb & Ws : see equation 6.10 (page 159)


6-26
Figure 6.6 Optimal Capital Allocation Line
for Bonds, Stocks and T-Bills

6-27
Figure 6.7 The Complete Portfolio

55% in O
45% in T-Bills

6-28
 Figure 6.8 The Complete Portfolio –
Solution to the Asset Allocation Problem
Optimal Portfolio (O):
Wb = 32.98%
Ws = 67.02%

Combination of riskless
asset (T-Bills) and
optimal portfolio (O) with
WT-bills = 45%:
Wb = 32.98% (.55*.3298)
Ws = 67.02% (.55*.6702)

6-29
6.4 EFFICIENT DIVERSIFICATION WITH
MANY RISKY ASSETS

6-30
Extending Concepts to All Securities

The optimal combinations result in lowest

level of risk for a given return.
The optimal trade-off is described as the
efficient frontier line (EFL).
These portfolios in EFL are dominant in
efficiency awareness.

6-31
Figure 6.9 Portfolios Constructed from
Three Stocks A, B and C

6-32
Figure 6.10 The Efficient Frontier of Risky
Assets and Individual Assets
Use different input data to
develop different efficient
frontiers: offers different
“optimal” portfolios.

Hence, real arena of the

competition among portfolio
managers is in the
sophisticated security analysis
that underlies their choices.

If quality of security analysis is

poor, a passive portfolio
(market index fund) can yield
better results than active
portfolios.
6-33
 Excel applications:
Efficient frontier for many stocks
Excel application: (Page 163)
INVESTMENT Course\Essentials of
investments_2010\Excel spreadsheets\chapter
6\Chapter_06_Efficient_Frontier_for_Many_Sto
cks
 Excel application:
INVESTMENT Course\Essentials of
investments_2010\Excel
spreadsheets\application_markovitz Model

6-34
 Efficient frontiers & a CAL for
diversification with or without short sale

Table 6.1: (page 165)

6-35
6.5 A SINGLE-FACTOR ASSET MARKET

6-36
Specification of a Single-Index Model of
Security Returns (1)
Assumption:
– One common factor is responsible for all the co-variability of stock
returns, with all other variability due to firm-specific factors.
– As usual, use the S&P 500 as a market proxy
Index model:
– Statistical models designed to estimate two components of risk for a
particular security or portfolio. It relates stock/portfolio return to
returns on both a broad market index & firm-specific influences.
Excess return can now be stated as:
Ri = βiRm + ei + αi (equation 6.11)
Single-variable regression equation of Ri on the market
excess return Rm.
6-37
Specification of a Single-Index Model of
Security Returns (2)
βi Rm
Component of return due to movements in the overall market (as
represented by the index Rm); βi is the security’s
responsiveness to the market factor.
ei
The component attributable to unexpected events that are
relevant only to this security (firm specific or residual risk).
αi
The stock’s expected excess return if the market factor is
neutral, that is, if the market’s excess return to zero (or a
constant value beyond any movements in the market index)

6-38
Figure 6.11 Scatter Diagram for Dell
Security characteristic line (SCL):

1. Plot of a security’s excess

return as a function of the
excess return of the market.

2. The line not represents the

ACTUAL returns, although the
actual returns are used to
calculate the regression
coefficient.

3. It represent average tendencies

(expectation), hence, it shows
the effects of the market return
on our expectation of RD.

E(RD|Rm) = αD + βD*Rm
6-39
 Components of Risk
Market or systematic risk: risk related to the macro
economic factor or market index
Unsystematic or firm specific risk: risk not related to the
macro factor or market index
Total risk = Systematic + Unsystematic
variance(Ri) = Variance(βiRm + ei + αi)
= Variance(βiRm) + Variance(ei)
i2 = i2 m2 + 2(ei)
Where:
i2 total variance
i2 m2 systematic variance
2(ei) unsystematic variance
6-40
Examining Percentage of Variance
the correlation coefficient b/t R i & Rm
2: proportion of total systematic (or explained) variance
of Ri that can be contributed to market fluctuations (R m)

2 = ßi2 m2 / 2
 2 = i2 m2 / [i2 m2 + 2(ei)]

• no firm-specific risk

• Large Abs(): systematic variance dominates the total
variance; that is, firm-specific variance is relatively
unimportant.

6-41
Figure 6.12 Various Scatter Diagrams

i2 = i2 m2 + 2(ei) 6-42

Advantages of the Single Index Model

High degree of applicability:

– Reduces the number of inputs for
diversification
– Easier for security analysts to specialize

6-43
Monitoring risk effects by setting up Security
characteristic line (SCL)
Example 6.3: (page 171)

βABC = Cov(RABC,Rmarket) / Var(Rmarket)

αABC = Average(RABC) – βABC*Average(Rmarket)

SCL : E(RD|Rm) = αD + βD*Rm

“Regression” approach:
Excel Filename: example 6.3

6-44
 Diversification
in a single-index security model
Beta of the portfolio = simple average of the individual
security betas:
>>> Any security in the portfolio is perfectly correlated
with the systematic part of any other security’s return, the
number of securities is of no consequence – Hence, no
diversification effects.
Firm-specific effects are independent of each other,
hence they are offsetting:
>>> But for unsystematic risk, by holding more securities,
even if has a large residual variance, sufficient
diversification can virtually eliminate firm-specific risks.

6-45
 Using security analysis w/t the index model:
01 benchmark passive portfolios & 01 active security
1. Passive portfolio (M): one or more index fund
2. ONE Active security (A): Google – D
3. Approach:
Compute benchmark line from index line & one security
(Google) instead of efficient frontier line, then
construct optimal portfolio (O) - w/t highest Sharpe
ratio.
WM & WA = (1 – WM)
computed by equation 6.10 w/t: A2 = A2 m2 + 2(eA)
4. S2O : computed from equation 6.17

6-46
 Using security analysis w/t the index model:
01 benchmark passive portfolios & 01 active portfolio
1. Passive portfolio (M): one or more index fund
2. ONE Active portfolio (A): Google – D & Dell - D
3. Approach:
(1) Compute optimal weight of each security in active portfolio:
Equation 6.18
(2) α, β of active portfolio, and residual variance is computed by
weighted average of each component security’s and weighted
sum of each security’s residual variance, respectively.
Equation 6.19
(3) Using Equation 6.15 & 6.16 & 6.17 to compute risk ratios for
optimal portfolio combined by active (A) & passive (M) portfolio.

6-47
Example 6.5: Treynor-Black Model
(page 174)

Table 6.2: Construction of optimal using

the index model

6-48
6.6 RISK OF LONG-TERM INVESTMENTS

6-49
 Are Stock Returns Less Risky
in the Long Run?
Consider a 2-year investment w/t stock returns in successive
years are almost uncorrelated (p=0 & common yearly σ):

Var (2-year total return) = Var (r1  r2

 Var (r1 )  Var (r2 )  2Cov(r1 , r2 )
  2  2  0
 2 2 and standard deviation of the return is  2
Variance of the 2-year return is double of that of the one-year
return and σ is higher by a multiple of the square root of 2.

6-50
Are Stock Returns Less Risky in the Long
Run?
Generalizing to an investment horizon of n years and
then annualizing:

Var(n-year total return) = n 2

Standard deviation (n-year total return) = n

1 
 (annualized for an n - year investment) =  n 
n n
Implications:
– Var(n-year total return) grows linearly with the number of years
– SD (n-year total return) grows in proportion to n
– This finding is true applicable to n – the number of stocks.

6-51
Are portfolio Return Less Risky w/t
more N components stocks ?
A portfolio of two identical, uncorrelated stocks (p=0)
having wa = wb = 0.5 & common yearly σ:
σ(portfolio) = σ / (2)1/2

Var(portfolio) = w2a*Var(a) + w2b*Var(b) + 2 wa* wb*Covar(a,b)

Conclusion:
– Portfolio risk over n component stocks declines. Hence,
diversification through rising a number of stocks in portfolio
is worth.

6-52
The Fly in the ‘Time Diversification’ Ointment

Annualized standard deviation is only appropriate

for short-term portfolios (not compounding risk
through time).

It can not serve to measure risk when comparing

investments of different horizons & different
scales of losses & returns.

6-53
The Fly in the ‘Time Diversification’ Ointment
To compare investments in two different
time periods, correctly based on:
– Risk of the total (end of horizon) rate of return,
which accounts for both magnitudes and
probabilities of possible losses.
– Insurance costs much more for longer
horizons.
A. One year in stock B. Two years in stocks
Possible outcome Probability Possible outcome Probability
Value doubles 0.5 Value quadruples 0.25
Value falls by half 0.5 Value unchanged 0.5
Value falls by 75% 0.25
6-54