FINA 3103

Intermediate Investments
HKUST

Spring 2025

Utpal Bhattacharya

Class 9: Index Models

© 2025 by Utpal Bhattacharya
All Rights Reserved
Selection of Securities Using Markowitz Model Is Impractical

• How Many Terms to Calculate?

n=5 σ11 σ12 σ13 σ14 σ15

variances
σ21 σ22 σ23 σ24 σ25
σ31 σ32 σ33 σ34 σ35
n(n-1)/2 = 10
unique σ41 σ42 σ43 σ44 σ45
covariances
σ51 σ52 σ53 σ54 σ55
© 2025 by Utpal Bhattacharya Slide 2
Index Models
All Rights Reserved
Markowitz Selection Impractical

 Too many estimates: n(n-1)/2 covariances

n = 5,000  12.5 million covariances…

 Each estimate has estimation error…

Average Returns >< Expected Returns
Sample Covariances >< True Covariances
… possibly causing big selection errors

© 2025 by Utpal Bhattacharya Slide 3

Index Models
All Rights Reserved
So What To Do: Make Two Simplifying Assumptions

 There exists a single macro factor that

• Summarizes all relevant macro data
• Moves the market as a whole
• Is the only source of returns correlation

 All other risk in a stock is firm specific

• No firm affects any other firm’s returns directly
© 2025 by Utpal Bhattacharya Slide 4
Index Models
All Rights Reserved
 What Do These Two Assumptions Mean?

 Your favorite stock goes up 10%. Why?

 We’re saying this realized return may reflect:

1) The fair expected return for that stock, E0(ri)

2) Surprise changes in the market as a whole, and

3) Surprise changes that affect this stock alone.

 More formally, we state these elements as:

© 2025 by Utpal Bhattacharya Slide 5

Index Models
All Rights Reserved
 Holding-Period (Realized) Returns

rit = E0(ri) + βi (rMt - E0(rM)) + eit

rit = Holding-period return for security i in period t

E0(ri) = Expected return for security i at time 0

βi (rMt - E0(rM)) = Impact of unanticipated systemic

events on I

= Surprise change of market X Effect of surprise change of market on i

eit = Impact of unanticipated firm-specific events on i

 E0 (rMt - E0(rM)) = E0(eit) = 0 by definition
© 2025 by Utpal Bhattacharya Slide 6
Index Models
All Rights Reserved
Single-Index Model

What is E0(ri)?
E0(ri) = αi + rf + βi(E0(rM) - rf)
This gives us the SI Model:
rit – rf = αi + βi(rMt – rf) + eit
intercept slope

rit – rf = Holding-period excess return for security i

αi = Expected excess return on i when excess return for market
index is zero (αi = 0 in equilibrium)
βi = Sensitivity of security i to market movements
rMt – rf = Excess return for market index in period t
eit = Impact of unanticipated firm-specific events on i

© 2025 by Utpal Bhattacharya Slide 7

Index Models
All Rights Reserved
Single-Index Model

rit – rf = αi + βi(rMt – rf) + eit

For your interest, I can rewrite it as
Basic return Market surprise

rit = αi + rf + βi(E0(rM) - rf) + βi(rMt - E0(rM)) + eit

which is expected return +

two unanticipated events, rMt - E0(rM) and eit

© 2025 by Utpal Bhattacharya Slide 8

Index Models
All Rights Reserved
Single-Index Model & Risk

systematic risk firm specific

Rit = αi + βiRMt + eit

where Rit = rit – rf and RMt = rMt – rf

The risk of a security is thus given by: Taking variance

σi 2 = βi σM
2 2 +σ 2(e )
i

Because σMe = 0 and σα = 0 (αi is a constant)

 i’s Total Risk = Systemic risk + Firm-specific risk

© 2025 by Utpal Bhattacharya Slide 9

Index Models
All Rights Reserved
Single-Index Model & Covariance
9 relationship

σij = Cov(Ri, Rj) = Cov(αi + βiRM + ei, αj + βjRM + ej)

= Cov(βiRM, βjRM)

σij = βi βjσM2
because σα = 0, Cov(RM, RM) = σM2 and σij(e) = 0

 It depends only on β & market risk.

 Only need to estimate n betas!
 Not n*(n-1)/2 covariances
© 2025 by Utpal Bhattacharya Slide 10
Index Models
All Rights Reserved
 Single-Index Model Example

 Ri = 0% + 0.9RM + ei , Rj = 0% + 1.1RM + ej
 σi(e) = 30%, σj(e) = 10%, and σM = 20% Statistical method

 Find σi , σj , and the covariance of i and j, σij

σi = [βi2σM2 + σ 2(ei)]1/2 = [(0.9)2(0.20)2+(0.30)2]1/2 = 35%

σj = [βj2σM2 + σ 2(ej)]1/2 = [(1.1)2(0.20)2+(0.10)2]1/2 = 24%

σij = βi βjσM2 + σij(e) = (0.9)(1.1)(0.20)2 + 0 = 3.96%

© 2025 by Utpal Bhattacharya Slide 11

Index Models
All Rights Reserved
Portfolio Alpha, Beta & Firm-Specific Risk

αP = Σwiαi

βP = Σwiβi

σ2(eP )= Σwi2σ2(ei)

© 2025 by Utpal Bhattacharya Slide 12

Index Models
All Rights Reserved
 The CAPM & The Index Model

 From the Index Model we can find the covariance of a security

with the market:
Market

 σiM = Cov(Ri, RM) = Cov(αi + βiRM + ei, RM)

= βiCov(RM, RM) + Cov(ei, RM)
Since Cov(ei, RM) = 0 and Cov(RM, RM) = σM2

Then: σiM = βiσM2  βi = σiM ÷ σM2

 Exactly what the CAPM tells us as well! covariance/market variance

© 2025 by Utpal Bhattacharya Slide 14

Index Models
All Rights Reserved
Discussion: Negative Beta Gold is opposite to stock performance (hedging)

 Can beta of a stock be negative? What

does it mean?

• E0(ri) = αi + rf + βi(E0(rM) - rf)

• If βi is negative, E0(ri) will be lower than

α i + rf

© 2025 by Utpal Bhattacharya Slide 15

Index Models
All Rights Reserved
Discussion: Negative Beta

 What does it mean?

• Negative beta means cov(ri, rM) is negative
• Roughly speaking, we get higher utility score
when we do well in down markets
• Negative beta assets are those that do
well in down markets Insurance
 People pay a high price to buy these
 Expected return is low
efficient frontier

hedges the portfolio in case of bad market

© 2025 by Utpal Bhattacharya Slide 16

Index Models
All Rights Reserved
Two Ways to Find Security Betas
(Download “Lecture 9 Index Model” from Canvas)

1. Market Variance & Security Covariance

βi = σiM ÷ σM2
Covariance/variance of another

where σiM & σM2 are calculated from sample returns

2. Regression Analysis of Sample Returns

(intercept) (slope) (residuals)

Rit = αi + βiRMt + eit

where the left-hand-side variable is Ri1 = rit – rf and the right-hand-

side variable is RMt = rMt – rf
© 2025 by Utpal Bhattacharya Slide 17
Index Models
All Rights Reserved
β from Variance & Covariance

Returns Variance-Covariance
IBM PG S&P
IBM 0.0094 -0.0003 0.0026
PG 0.0072 0.0008 covariance/market variance
S&P 0.0021

Betas
2
β IBM 1.2248 σ IMB,S&P / σ S&P
2
β PG 0.3624 σPG,S&P / σS&P
2
β S&P 1.0000 σ S&P,S&P / σ S&P

© 2025 by Utpal Bhattacharya Slide 18

Index Models
All Rights Reserved
 β from Regression Analysis
rit – rf = αi + βi(rMt – rf) + eit
0.30
IBM's Actual Excess Return
0.25
R2 = systematic risk/Total Risk
IBM's Predicted Excess Return = systematic risk/(systematic risk+Firm risk)
0.20
0.15 Positive Residual,
α: Intercept
0.10 eit = rit – E(ri) > 0
0.05 β: Slope
Market sensitivity
0.00
-0.05
-0.10 R2: strength of association
deviation from CAPM
-0.15 t-stats: more than 2, statistically significant
Negative Residual, p-value: 5% or below
-0.20
eit = rit – E(ri) < 0
-0.25
S&P Excess Return
-0.10 -0.08 -0.06 -0.04 -0.02 0.00 0.02 0.04 0.06 0.08 0.10
© 2025 by Utpal Bhattacharya Slide 19
Index Models
All Rights Reserved