Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Index Models
Juan Sotes-Paladino
FNCE30001
Investments
Semester 2 2015
University of Melbourne FBE
Index Models
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Overview
Motivation
Todays Class...
A Single Index Model for Security
Returns
A solution to drawbacks of the
Markowitz model
An application of the ideas in
the CAPM
Diversification under the Index
Model
The Amsterdam Stock Exchange, now called Euronext Amsterdam,
is considered the oldest stock exchange in the world. It was
established in 1602 by the Dutch East India Company to trade its
stocks and bonds. Too bad we do not have good data from it
Index Models in Practice
Empirical Validity
testing the CAPM might be easier.
University of Melbourne FBE
Index Models
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Overview
Motivation
A Motivation: Drawbacks of Markowitz Model
In order to implement the model with N assets, a large number of covariances is
required
One for each unique pair of assets
Example: a portfolio of 50 securities requires:
N = 50 estimates of expected returns
N = 50 estimates of variances
N (N 1)/2 = (50 49)/2 = 1, 225 estimates of covariances
Total: 1,325 estimates
Real-world applications can get even worse:
Australia: approx. 2,200 unique stocks 2.4 million covariance computations
(approx.)
U.S.: approx. 10,000 unique stocks 50 million covariance computations (approx.)
University of Melbourne FBE
Index Models
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Overview
Motivation
How Do We Estimate these Covariances?
Why is the large number of covariances a problem?
After all, todays computers can perform complex calculations in no time
Main problem is that estimation error can make Markowitz model almost useless
Remember: covariances most likely estimated from past data
Estimates will look like confidence intervals: e.g.,
BHP,QANTAS [0.346, 0.592]
wBHP
and wQANTAS
can vary a lot depending on the value of
BHP,QANTAS in
[0.346, 0.592] we use!
Problem can get so serious that simply equal-weighting the assets (w = 1/N) in a
portfolio can do much better
DeMiguel, Garlappi & Uppal, 2007, Optimal Versus Naive Diversification: How
Inefficient is the 1/N Portfolio Strategy? Review of Financial Studies.
University of Melbourne FBE
Index Models
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Overview
Motivation
A Solution
Q: How can we reduce the number of inputs needed to estimate covariances?
By reducing the number of inputs, we hope to:
reduce calculations
limit estimation errors
A: Use an index model of asset returns
A special case of a factor model
where the factor is a market index
as suggested by the CAPM
University of Melbourne FBE
Index Models
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Single Factor Model
Single Index Model
Single Index Model
University of Melbourne FBE
Index Models
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Single Factor Model
Single Index Model
The Basics: a Single Factor Model
A factor model is a statistical representation of the distinction between systematic
and firm-specific risks of a security:
Systematic risk is largely macroeconomic and affects all securities
Firm-specific affects the security only
Start by expressing the rate of return on a security as the sum of its expected and
unexpected parts:
ri = E (ri ) + ui
where ui measures the unexpected component of security returns, with:
E (ui ) = 0
Var (ui ) = i2
e.g., if the expected returns was 5% and the actual return turned out to be 7%, the
unexpected component was 2%.
This is OK; we can always decompose a random variable like this
If not convinced, take expectations on both sides of the equation
University of Melbourne FBE
Index Models
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Single Factor Model
Single Index Model
A Common Factor
The returns on jointly normally distributed securities are linearly related to one or
more common factors
A macroeconomic variable, denoted by m, that captures unanticipated macro
surprises:
2
E (m) = 0, var (m) = m
e.g., unemployment rate 0.2% higher than expected
In this case, we can further decompose uncertainty ui into uncertainty about the
economy as a whole (m) and uncertainty about the firm ei
ri = E (ri ) + i m + ei
| {z }
ui
where i is a is sensitivity coefficient to macroeconomic conditions
We call this equation a single factor model
University of Melbourne FBE
Index Models
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Single Factor Model
Single Index Model
Variances and Covariances in the Factor Model
Properties of m:
Has no subscript, because it affects all securities
Not correlated with firm specific risk: for all i, cov (m, ei ) = 0
The variance of ri is, then:
2
i2 = i2 m
+ 2 (ei )
Since all securities are affected by m and firm-specific uncertainty (ei ) is
uncorrelated across securities (cov (ei , ej ) = 0 for i 6= j), the covariance between
any two securities is:
2
cov (ri , rj ) = cov (i m + ei , j m + ej ) = i j m
University of Melbourne FBE
Index Models
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Single Factor Model
Single Index Model
What is m? The Single Index Model
The CAPM has an obvious suggestion for the choice of m:
Use (a proxy for) the market portfolio: a market index
Each major market has at least one index we could use for m:
Australia: All Ordinaries, S&P/ASX 200
US: S&P 500, Dow Jones
London: FTSE 100
Japan: Nikkei 225
Germany: DAX
University of Melbourne FBE
Index Models
10
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Single Factor Model
Single Index Model
How do we estimate ?
We need to estimate systematic risk, beta:
Estimations of i are key to estimate variances and covariances
Easy to do when we have past data for the index
Let us define
Ri ri rf
RM rM rf
Ri (t) = i + i RM (t) + ei (t)
The Single Index Model
Defines a security characteristic line (SCL) for asset i: Regression line of best fit
through a scatterplot of rates of return for an individual risky asset (i) and for the
market portfolio of risky assets (M) over some designated past period
University of Melbourne FBE
Index Models
11
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Single Factor Model
Single Index Model
Estimating the SCL
Ri (t) = i + i RM (t) + ei (t)
We estimate the single index model by collecting historical data for Ri and RM
The intercept is the expected excess return of the security (when the market
excess return is zero)
A nonmarket premium
Securitys beta is the securitys sensitivity to the index
ei is the error term (the firm-specific residual)
University of Melbourne FBE
Index Models
12
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Single Factor Model
Single Index Model
Risk Premium
Taking the expected return of the regression equation, we have:
E (Ri ) = i + i E (RM )
Looks familiar?
Part of a securitys risk premium E (Ri ) is from the risk premium of the index
E (RM ): i E (RM )
According to the CAPM, this should be the only risk premium
The rest of the risk premium is firm-specific, denoted by i
According to the CAPM, i =?
Fund managers try to find securities with non-zero alpha
University of Melbourne FBE
Index Models
13
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Diversification
University of Melbourne FBE
Index Models
14
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
The Single Index Model and Diversification
Investors can also achieve diversification with the single index model
Firm-specific risk can be eliminated whereas market (systematic) risk remains
Suppose we choose an equally-weighted portfolio P of N securities
RP = P + P RM + eP
P =
N
1 X
i
N i=1
P =
N
1 X
i
N i=1
eP =
N
1 X
ei
N i=1
2
P2 = P2 m
+ 2 (eP )
University of Melbourne FBE
Index Models
15
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Average Idiosyncratic Risk
Let us define the average idiosyncratic risk of the components of P as
P2 =
1
N
PN
i=1
2 (ei )
For a sufficiently large number N, this average should remain relatively constant
as we add more stocks to the portfolio
Empirically,
P2 for N = 50 generally very close to
P2 for N = 100
Lets go back to the residual term eP in our portfolio:
E [eP ] =
N
1 X
E [ei ] = 0
N i=1
P = var
N
N
1 X
X
1
ei = 2 var
ei
N i=1
N
i=1
N
1 X 2
(ei )
2
N i=1
1 2
N P
University of Melbourne FBE
(since cov (ei , ej ) = 0)
Index Models
16
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Limits to Diversification
As N increases,
P2 is unchanged but 1/N approaches 0 quickly
Thus, as N increases, 2 (eP ) =
1 2
(e)
N
University of Melbourne FBE
Index Models
17
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Index Models in Practice
University of Melbourne FBE
Index Models
18
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Index Models and Security Analysis
According to the index model, the risk premium of a stock is derived from the
stocks tendency to follow the market index ()
Any expected return beyond the market risk premium () comes from a
non-market factor
The goal of security analysis
This distinction provides a natural division of labour across analysts
Market timers: macroeconomic analysts that estimate risk premium E (RM ) and
2 of the market index
risk M
Security analysts: provide the security-specific expected return forecasts/estimates
(alphas i ) from security valuation models
Statistical analysis can be used to estimate betas i and residual variances 2 (ei )
University of Melbourne FBE
Index Models
19
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Security Analysis and Portfolio Construction
Security analysis will uncover stocks like A and B:
E(r)
Market
rf
NOTE: for security analysis to work, we need market to be at most nearly
efficient
If fully efficient, we go back to CAPM
All alphas should be zero!
If we trust our security analysis, how should we invest in A and B?
University of Melbourne FBE
Index Models
20
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Optimal Risky Portfolios and the Single-Index Model
If we believe a single index model holds, we can greatly simplify the process of
constructing optimal risky portfolios
The basic tradeoff is:
If interested only in diversification, just hold the market portfolio
No extra reward (alpha) on any security
If interested in enhancing portfolio reward-to-risk, take a differential position in
securities with nonzero alpha
Implies assuming unnecessary firm-specific risk
Treynor & Black (73) solution: The optimal risky portfolio is a combination of:
An active portfolio, denoted by A: comprises the n analyzed securities
Follows from security analysis
The market index (passive) portfolio, denoted by M: the (n + 1)-th asset included
to aid in diversification
University of Melbourne FBE
Index Models
21
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
The Treynor& Black Approach: An Overview
Treynor& Black show that the Sharpe ratio of the optimal risky portfolio is
2
A
2
SP2 = SM
+
(eA )
The (squared) Sharpe ratio of the optimal risky portfolio is higher than that of the
i2
h
A
index portfolio only by (e
)
A
The ratio of a portfolios alpha to its residual standard deviation is called the
A
information ratio, (e
A)
Extra return we can obtain from security analysis compared to the firm-specific risk
we incur when we over- or underweight securities relative to the passive market index
The intuition is then that, to maximize the Sharpe ratio of the risky portfolio, the
information ratio of the active portfolio needs to be maximized
Achieved by investing in each security in proportion to its ratio of
i
2 (ei )
In turn, the weight of the active portfolio in the optimal overall risky portfolio P is
proportional to its ratio of 2(eA )
A
University of Melbourne FBE
Index Models
22
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
The Input List: Comparison with Markowitz Model
The Treynor& Black approach, based on the single index model, only requires the
following estimates for portfolio optimization:
n estimates of the extra-market expected excess returns, i
n estimates of beta, i
n estimates of firm-specific variances, 2 (ei )
One estimate for the market risk premium, E (RM )
2
One estimate for the variance of the market component, M
A total of (3n + 2) estimates
For a 50 security portfolio, this means 152 estimates are needed
As opposed to the earlier 1,325 estimates
Allows for specialization of effort in security analysis
e.g., specialization by industry: aerospace, pharmaceutical, technology, financial
Difficult otherwise for covariance terms involving firms in different industries
University of Melbourne FBE
Index Models
23
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
The Optimization Procedure
Once the input list is complete, the optimal risky portfolio can be formed as follows:
1. Compute the initial position of each security in the active portfolio as:
i
wi0 = 2
(ei )
2. Scale the initial positions to make portfolio weights sum to 1 by dividing by their
sum:
w0
wi = PN i 0
i=1 wi
3. Compute the alpha of the active portfolio
A =
N
X
wi i
i=1
4. Compute the residual variance of the active portfolio
2 (eA ) =
N
X
wi2 2 (ei )
i=1
University of Melbourne FBE
Index Models
24
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
The Optimization Procedure (contd)
5. Compute the initial position in the active portfolio
wA0 =
A / 2 (eA )
2
E (RM )/M
6. Compute the beta of the active portfolio
A =
N
X
wi i
i=1
7. Adjust the initial position in the active portfolio
wA =
wA0
1 + (1 A )wA0
University of Melbourne FBE
Index Models
25
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
The Optimization Procedure (contd)
8. Find the weight of the passive portfolio
wM
= 1 wA
9. Calculate the risk premium of the optimal risky portfolio from the risk premium of
the index portfolio and the alpha of the active portfolio
E (RP ) = (wM
+ wA A )E (RM ) + wA A
10. Compute the variance of the optimal risk portfolio from the variance of the index
portfolio and the residual variance of the active portfolio
2
P2 = (wM
+ wA A )2 M
+ (wA (eA ))2
University of Melbourne FBE
Index Models
26
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Example: An Optimal Portfolio
Consider the following information for four stocks; A, B, C and D:
Stock
A
B
C
D
Expected Return
0.30
0.25
0.12
0.25
Beta
1.9
1.2
1.6
0.7
Residual SD
0.45
0.49
0.38
0.22
Suppose that the risk-free rate is 4% and a passive equity portfolio (an index
portfolio) has a 15% expected return with a 15% standard deviation
Optimal risky portfolio?
University of Melbourne FBE
Index Models
27
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Example (contd)
We first find alphas:
E (ri ) rf = i + i (rM rf )
i = E (ri ) (rf + i (rM rf ))
So
A
B
C
D
= 0.30 (0.04 + 1.9(0.15 0.04)) = 0.051
= 0.25 (0.04 + 1.2(0.15 0.04)) = 0.078
= 0.12 (0.04 + 1.6(0.15 0.04)) = 0.096
= 0.25 (0.04 + 0.7(0.15 0.04)) = 0.133
and the residual variances are
2 (eA ) = 0.452 = 0.2025
2 (eB ) = 0.492 = 0.2401
2 (eC ) = 0.382 = 0.1444
2 (eD ) = 0.222 = 0.0484
University of Melbourne FBE
Index Models
28
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Example (contd)
Step 1: Find the initial positions of each security in the portfolio
Step 2: Scale those weights to make portfolio weights equal to 1
Stock
i
A
B
C
D
Total
Step 1
wi0
i
2 (ei )
0.2519
0.3249
-0.6648
2.7479
2.6598
University of Melbourne FBE
Step 2
wi =
w0
PN i
i=1
wi0
0.0947
0.1221
-0.2499
1.0331
1.0000
Index Models
29
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Example (contd)
Step 3: Compute the alpha of the active portfolio
A =0.0947 0.051 + 0.1221 0.078 + (0.2499) (0.096) + . . .
. . . + 1.0331 0.133 = 0.1758
Step 4: Compute the residual variance of the active portfolio
A2 =0.09472 0.2025 + 0.12212 0.2401 + (0.2499)2 (0.1444) + . . .
. . . + 1.03312 0.0484 = 0.0661
University of Melbourne FBE
Index Models
30
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Example (contd)
Step 5: Compute the initial position in the active portfolio
wA0 =
A / 2 (eA )
0.1758/0.0661
=
= 0.5441
2
(0.15 0.04)/0.0225
E (RM )/M
Step 6: Compute the beta of the active portfolio
A =
N
X
wi i
i=1
=0.0947 1.9 + 0.1221 1.2 + (0.2499) 1.6 + 1.0331 0.7 = 0.6497
University of Melbourne FBE
Index Models
31
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Example (contd)
Step 7: Adjust the initial position in the active portfolio
wA =
wA0
0.5441
= 0.4570
=
1 + (1 0.6497)(0.5441)
1 + (1 A )wA0
Step 8: Find the weights of the active portfolio and the index portfolio in the
optimal risky portfolio
wM
= 1 wA = 1 0.4570 = 0.5430
University of Melbourne FBE
Index Models
32
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Example (contd)
Step 9: Calculate the risk premium of the optimal risky portfolio
E (RP ) =(wM
+ wA A )E (RM ) + wA A
=(0.5430 + 0.4570 0.6497)(0.15 0.04) + 0.4570 0.1758 = 0.1727
Step 10: Compute the variance of the optimal risky portfolio
2
P2 =(wM
+ wA A )2 M
+ (wA (eA ))2
=(0.5430 + 0.4570 0.6497)2 0.0225 + (0.4570
0.0661)2 = 0.0297
Sharpe ratio comparison
SP =
0.1727
0.0297
= 1.0021
University of Melbourne FBE
SM =
Index Models
0.11
0.0225
= 0.7333
33
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Example Use of the Single Index Model
Portfolio of 6 stocks (+ the market portfolio): H-P, Dell, Target, Wal-Mart, BP,
and Shell
For each stock, we can get estimates of betas i and residual variances 2 (ei )
from estimating their SCL:
Obtain 60 monthly observations of rates of return for each stock and the S&P 500
index
Compute the excess returns (over T-bills) on the 7 components of the portfolio
For each individual stock (e.g., HP), estimate the regression equation:
RHP (t) = HP + HP RM (t) + eHP (t)
University of Melbourne FBE
Index Models
34
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
SCL: Scatter Diagram
University of Melbourne FBE
Index Models
35
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
SCL: Regression Statistics
Example: The Single Index Model
52% of the variation in
the HP is explained by
the variation in the S&P
500 excess returns
Portion of the variance
of the HP s excess
return explained by the
S&P 500 excess return
Monthly
standard
deviation of
HP s
residuals
alpha
Beta
17
How can we use the regression output to infer relevant quantities?
Standard error = .0767 2 (eHP ) = .07672 = .0059 (compare with MS)
R2 =
2
2
HP
m
2
P
R2 =
2
2
HP
m
2
P
=1
2 (eHP )
2
P
2 =
m
P2 =
2
R P
2
HP
2 (eHP )
1R 2
.52390.0124
2.03482
University of Melbourne FBE
.0059
1.5239
= 0.0124
= 0.0016
Index Models
36
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Is the Index Model Inferior to the Markowitz Model?
Relative to the Markowitz Model:
The single index model places more restrictions on the structure of asset return
uncertainty
The classification of uncertainty into macro versus firm-risk oversimplifies the
real-world uncertainty
e.g., industry-risk is ignored
Thus, the optimal portfolio derived from the single-index model might be inferior
to the one obtained from the Markowitz model
University of Melbourne FBE
Index Models
37
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Active Management
Implementation
Comparison with Markowitz
Is the Index Model Inferior to the Markowitz Model? (cont.)
The Efficient Frontier: Single Index vs. Markowitz
For our 6-stock portfolio, the difference between the two approaches is negligible:
University of Melbourne FBE
Index Models
38
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Validity of the CAPM
Violations of the CAPM
Potential Explanations and Remedies
Empirical Validity of the Index Model
University of Melbourne FBE
Index Models
39
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Validity of the CAPM
Violations of the CAPM
Potential Explanations and Remedies
Relation to the CAPM
We can rewrite the index model as:
E (ri ) rf = i + i (E (rm ) rf )
The CAPM is:
E (ri ) rf = i (E (rm ) rf )
We can show that the beta in the index model equals the beta in the CAPM
However, implications for alpha differ
The CAPM implies i = 0, in expectation, for each security
By contrast, the index model implies that the realized value of alpha should be zero
on average across securities
P
P
i.e.
i wi i = 0, since
i wi ri = rM
University of Melbourne FBE
Index Models
40
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Validity of the CAPM
Violations of the CAPM
Potential Explanations and Remedies
Is the CAPM Valid?
The single index model is closely related to the CAPM
Thus, its practical relevance depends to a great extent on the CAPMs empirical
validity
So...how does the CAPM fare with the data?
While its implications are qualitatively supported
empirical tests do not support its quantitative predictions
Note: the CAPM is derived using expected returns, which are not observed
University of Melbourne FBE
Index Models
41
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Validity of the CAPM
Violations of the CAPM
Potential Explanations and Remedies
Empirical Tests of the CAPM
Estimating the SCL gets us alpha and beta estimates for a given stock.
It is true that if CAPM holds, we should find an alpha estimate of zero (and
practically speaking, a beta estimate between 0.5 and 2.0).
We could take these betas for a number of assets or portfolios, and use them as
x-variables in a cross-sectional test of CAPM:
rit rf = a + b i + it .
Does it work?
University of Melbourne FBE
Index Models
42
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Validity of the CAPM
Violations of the CAPM
Potential Explanations and Remedies
CAPM Tests: Initial Success
Initially, the CAPM was viewed as successfully capturing a positive relationship
between expected return and betas.
CAPMMean excess returns vs. beta, version 1
mean excess returns, percent
18
Means and betas
Fitted market premium
Direct market premium
14
10
2
0.0
0.2
0.4
0.6
0.8
betas
1.0
1.2
1.4
Notes: Average returns versus betas on the NYSE value-weighted
portfolio for ten size-sorted stock portfolios, government bonds,
and corporate bonds. Sample period 194796. The black line
draws the CAPM prediction by fitting the market proxy and
Treasury bill rates exactly (a time-series test) and the colored line
draws the CAPM prediction by fitting an OLS cross-sectional
regression to the displayed data points (a second-pass or crosssectional test). The small-firm portfolios are at the top right.
Moving down and to the left, one sees increasingly large-firm
portfolios and the market index. The points far down and to the
left are the government bond and Treasury bill returns.
University of Melbourne FBE
Index Models
43
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Validity of the CAPM
Violations of the CAPM
Potential Explanations and Remedies
Violations of the CAPM
Later tests of the CAPM did not support its validity, though:
Cross-sectional violations: even after accounting for beta
the market capitalization of a firm is a predictor of its average historical return (size
effect)
Stocks with low market-to-book ratios tend to have higher returns than stocks with
high market-to-book ratios (value effect)
Stocks that have performed well over the past 6 months tend to have high expected
returns over the following six months (momentum)
Time-series violations: even after controlling for beta
Firms with high P/E ratios have lower return
Stocks with high dividend yield have higher returns
University of Melbourne FBE
Index Models
44
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Validity of the CAPM
Violations of the CAPM
Potential Explanations and Remedies
Why Does the CAPM Fail?
The CAPM may fail to capture other risks:
Distress Risk. Value stocks tend to be stocks that have underperformed in the past.
They could be in financial distress making them riskier.
Liquidity Risk. Small stocks are more illiquid and may thus command a higher
premium
The proxies for the market portfolio do not fully capture all of the relevant risk
factors in the economy
For example, human capital is excluded from the various proxies (ASX200) for the
market portfolio
E.g., large firms may be perceived to be less vulnerable to economic downturns that
diminish the value of human capital lower risk-premium
There may be behavioral biases against classes of stocks, which have nothing to
do with the reward-to-risk ratios on stocks
E.g., portfolio managers dont lose their jobs for investing in BHP
...but they may if they invest in a financially distressed company when it is selling
for $0.10/share
University of Melbourne FBE
Index Models
45
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Validity of the CAPM
Violations of the CAPM
Potential Explanations and Remedies
Potential Remedies
Despite its shortcomings, the CAPM is widely employed
Most recent research admits multiple factors as determinants of risk
This is the subject of next seminar!
University of Melbourne FBE
Index Models
46
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Appendix: Size Effect
Appendix: Value Effect
Appendix: Momentum Effect
Takeaways
A single-factor model of the economy classifies sources of uncertainty as
systematic (macroeconomic) factors or firm-specific (microeconomic) factors.
The index model assumes that the macro factor can be represented by a broad
index of stock returns.
It drastically reduces the necessary inputs in the Markowitz approach.
Estimated by applying regression analysis to excess rates of return.
Optimal active portfolios constructed from the index model include analysed
securities in proportion to their information ratios.
The empirical failure of the CAPM suggests that more than one risk factor are
necessary to explain the cross-section of stock returns.
To be continued next seminar...
University of Melbourne FBE
Index Models
47
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Appendix: Size Effect
Appendix: Value Effect
Appendix: Momentum Effect
Additional Resources
Plenty of online resources for you to explore valuation and portfolio selection
issues:
Value Line Investment survey: popular source of finding a stocks beta, but also
investment-related articles
http://www.valueline.com/
William Sharpes website: the 1990 Nobel prize in Economics winner has plenty of
articles on asset valuation and allocation, applications to retirement saving, etc.
http://www.wsharpe.com/
http://www.moneychimp.com/: informative education site on investments that
includes CAPM calculators for estimating a stocks return and a market simulator
University of Melbourne FBE
Index Models
48
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Appendix: Size Effect
Appendix: Value Effect
Appendix: Momentum Effect
Appendix
University of Melbourne FBE
Index Models
49
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Appendix: Size Effect
Appendix: Value Effect
Appendix: Momentum Effect
The Small Firm Effect
Size Effect A number of researchers, starting with Keim (1981) and Banz
(1981) found that differences in firm size explained differences in
expected
returns.
Researchers
such as
Keim (1981) and Banz (1981) found that firm size, defined
by total market
also
differences
in expected returns.
1. firmcapitalization,
size was defined
as explained
total market
capitalization.
2. Small stocks (i.e. small cap stocks) outperformed large stocks (i.e.
In particular,
small cap stocks tended to outperform large cap stocks in the data.
large cap stocks).
To get aInhandle
onminimize
this effect,
we can builderror,
portfolios
sorted
past market
order to
measurement
we will
formon
portfolios
of
capitalizations.
stocks based on their past market capitalizations
Dec
MKCap(m$)
NYSE
10
9
8
7
6
5
4
3
2
1
511,391
10,486
4,428
2,237
1,387
889
534
353
198
95
172
172
172
172
172
172
172
172
172
172
FIN460-Papanikolaou
University of Melbourne FBE
# of Stocks
AMEX
NASDAQ
5
3
5
5
5
11
15
32
73
412
80
81
136
166
217
254
251
400
551
1,399
CAPM and empirical evidence
Index Models
Total
257
256
313
343
394
437
438
604
796
1,983
14/ 37
50
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Appendix: Size Effect
Appendix: Value Effect
Appendix: Momentum Effect
The Small Firm Effect
Size Effect - Returns
Excess portfolio returns, historical average (1926-2007)
Excess portfolio returns on market cap sorted portfolios from 1926 to 2007:
16
14
Excess return, % (historical average)
12
10
0
1
10
MKCAP Decile
Consistent with the CAPM?
FIN460-Papanikolaou
University of Melbourne FBE
CAPM and empirical evidence
Index Models
15/ 37
51
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
The Small Firm Effect
Appendix: Size Effect
Appendix: Value Effect
Appendix: Momentum Effect
Size Effect To
- Size
Sorts
resolve this, Fama and French (1992) do a double sort, first on
size, then on market b.
Fama & French (1992) do a double sort, first on size and then on market .
1/Size ("Smallness")
Market Beta
Should the point cloud be that
thick to be consistent
with the CAPM?17/ 37
FIN460-Papanikolaou
CAPM and empirical evidence
University of Melbourne FBE
Index Models
52
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Appendix: Size Effect
Appendix: Value Effect
Appendix: Momentum Effect
Size Effect - Fama & French
Fama & French (1992) find that the relation between average returns and within a
size decile is generally negative.
University of Melbourne FBE
Index Models
53
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Appendix: Size Effect
Appendix: Value Effect
Appendix: Momentum Effect
The Value Effect
Way back in the 1930s, Graham and Dodd noticed that value stocks tend to
outperform growth stocks on average.
Definition: A value stock is a stock with a low market price relative to the book
value of assets.
Definition: A growth stock is a stock with a high market price relative to the
book value of assets.
Value stocks are characterized by some as being undervalued by the market,
while growth stocks are characterized as being glamour stocks that are
overvalued.
Later work argues that these two types of stocks simply have different risk
characteristics.
University of Melbourne FBE
Index Models
54
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Appendix: Size Effect
Appendix: Value Effect
Appendix: Momentum Effect
Value vs. Beta
Suppose
weEffect
sort all stocks into 10 portfolios based on the ratio of the book value
The
Value
of equity to the market value of equity with rebalancing occurring yearly.
Here Puzzle
are the
characteristics
portfolios
is more
pronouncedof
in these
the post-war
periodin the post-war era:
Sort
Lo
3.95
(2.67)
18.01
2
5.59
(2.42)
16.35
a
(t)
bMKT
(t)
R2 (%)
-2.07
(1.05)
1.10
(0.02)
86.02
-0.04
(0.73)
1.03
(0.02)
91.34
E(Ri )
rf
10 portfolios sorted on book-to-market equity, 1962-2007
3
4
5
6
7
8
6.07
6.29
6.28
7.28
8.33
8.67
(2.39)
(2.37)
(2.22)
(2.21)
(2.21)
(2.19)
16.14
16.00
15.01
14.90
14.92
14.78
0.54
(0.76)
1.01
(0.02)
90.27
0.95
(0.95)
0.97
(0.03)
85.63
1.38
(1.01)
0.89
(0.03)
81.92
University of Melbourne FBE
2.38
(0.92)
0.90
(0.03)
83.31
3.63
(1.10)
0.86
(0.03)
76.58
Index Models
4.05
(1.12)
0.84
(0.03)
75.20
9
9.64
(2.37)
16.01
Hi
11.11
(2.73)
18.46
Hi-Lo
7.16
(2.28)
15.39
4.66
(1.23)
0.91
(0.04)
74.27
5.71
(1.71)
0.99
(0.05)
65.82
7.78
(2.44)
-0.11
(0.06)
1.07
55
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Appendix: Size Effect
Appendix: Value Effect
Appendix: Momentum Effect
Size & Value Together
Size and Value Effect together
25 Portfolios Sorted on Size & Book-to-Market
25 portfolios sorted on Size and Book to Market
Average Excess Portfolio
Returns vs. from 1962 to 2007
Average excess portfolio returns vs market beta (1962-2007)
18
Small Value
Average excess return (%)
16
14
12
Large Value
10
8
Large Growth
6
Small Growth
4
2
0.7
0.8
0.9
1.1
1.2
1.3
1.4
1.5
1.6
Market beta
FIN460-Papanikolaou
University of Melbourne FBE
CAPM and empirical evidence
Index Models
23/ 37
56
Intro
Single Index Model
Diversification
Index Models in Practice
Empirical Validity of the Index Model
Conclusions
Appendix: Size Effect
Appendix: Value Effect
Appendix: Momentum Effect
The
MomentumMomentum
Effect
Short-Term
Jagadeesh & Titman (1993) showed that firms with high (low) returns in the
previous
tendoftostocks
haveselected
high (low)
returns
in the
following
few
months.
Form year
portfolios
on their
past return
over
the last 12
months.
will rebalance these portfolios every month.
FormWeportfolios
based on past performance over the last 12 months with monthly
rebalancing.
Lo
0.31
(3.70)
33.24
10 portfolios sorted on previous 12 month returns (1926-2007)
2
3
4
5
6
7
8
9
5.18
5.12
6.72
6.80
7.58
8.68
10.30
11.47
(3.13)
(2.70)
(2.50)
(2.31)
(2.26)
(2.17)
(2.09)
(2.20)
28.15
24.24
22.41
20.77
20.33
19.51
18.78
19.78
Wi
15.37
(2.52)
22.64
Wi-Lo
15.07
(2.94)
26.44
a
(t)
bMKT
(t)
R2 (%)
-11.52
(1.65)
1.53
(0.08)
74.95
-5.08
(1.31)
1.33
(0.07)
78.61
7.48
(1.33)
1.02
(0.06)
71.90
19.01
(2.44)
-0.51
(0.13)
13.05
Sort
E(Ri )
rf
-3.91
(1.12)
1.17
(0.06)
82.24
-1.77
(0.94)
1.10
(0.04)
85.01
-1.18
(0.79)
1.03
(0.04)
87.30
-0.39
(0.66)
1.03
(0.02)
90.93
1.09
(0.70)
0.98
(0.02)
89.64
3.05
(0.70)
0.94
(0.02)
88.34
3.97
(0.81)
0.97
(0.03)
85.09
Again, this is quite inconsistent with the CAPM. Momentum seems to also exist in
many different assets classes such as commodities and currencies.
University of Melbourne FBE
Index Models
57
