20203A

Portfolio management
1

Class 7
APT & Multifactor models

© Melanie Harris 2022 (updated in Oct 2023) Source: Jean-Philippe Tarte.

 Outline class 7
2

Section 1: Introduction Law of One Price

Section 2: APT model
Section 3: Factor models
 Section 1: Introduction.
General form of factor models of security returns
3

 Recall stock variability may be decomposed into the following sources:

 Market (i.e., systemic) risk
 Largely due to macroeconomic events
 Firm-specific (i.e., idiosyncratic) effects
Multiple risk factors.
𝑅𝑖 = 𝐸 𝑅𝑖 + 𝑏𝑖1𝜆1 + 𝑏𝑖2𝜆2 + ⋯ + 𝑏𝑖𝑘 𝜆𝑘 + 𝜀𝑖 for 𝑖 = 1 to 𝑛
where:
Ri = actual excess return on asset i during a specified time period, i = 1, 2, 3, … n. n is number of assets
E(Ri) = expected excess return for asset i (if all the risk factors have zero changes)
bij = Factor surprise sensitivity or factor loading or factor beta. Sensitivity of asset i’s returns to movements in the
common risk factor j
𝜆k = surprise in common factor k (𝜆 k could be positive or negative but has an expected value of zero and influences the
returns on all assets). k number of factors.
εi = firm-specific surprise events of asset i unrelated to the macro factors (a random error term that, by assumption, is
completely diversifiable in large portfolios and has a mean of zero)

The general form is a description of the factors that affect security returns… Where
𝐸 𝑅𝑖 comes from? We need a theoretical model of equilibrium security returns to help
determine the expected value 𝐸 𝑅𝑖 …
 Section 2: Arbitrage pricing theory (APT model)
4

A. Assumptions and equation

B. Comparing the CAPM and the APT
C. Using the APT: Two-stock and a two-factor model example
 A – Assumptions and equation
5

 APT developed by Ross (1976). An alternative to CAPM. A multifactor model.

𝑅𝑖 = 𝐸 𝑅𝑖 + 𝑏𝑖1 𝜆1 + 𝑏𝑖2 𝜆2 + 𝜀𝑖 and 𝐸 𝑅𝑖 = 𝑏𝑖1 𝐸(𝜆1 ) + 𝑏𝑖2 𝐸(𝜆2 ) 2 factor SML.

 APT is intuitive and makes fewer assumptions than CAPM. Equilibrium model.
𝑅𝑖 = 𝐸 𝑅𝑖 + 𝑏𝑖 𝐹 + 𝜀𝑖 and 𝐸 𝑅𝑖 = 𝑏𝑖 𝐸 𝑅𝑚 R = excess return 1 factor SML.

 Three major assumptions:

 A factor model describes asset returns
 With many assets to choose from, investors can form well-diversified portfolios that eliminate asset-
specific risk
 No arbitrage opportunities exist among well-diversified portfolios.

 In contrast to CAPM, APT does not assume:

 Normally distributed security returns
 Quadratic utility function
 A mean-variance efficient market portfolio
In the case of AOA, the APT says something very similar to the CAPM
 Assumptions and equation
6

 APT is a return-generating process that can be represented as a K factor model

 Assuming that a risk-free investment yields rf, we find that in equilibrium, the expected return
on any asset i can be expressed as:

Risk premium related

to the jth common
risk factor

Fundamental
𝐸(𝑟𝑖) = 𝑟𝑓 + 𝜆1𝑏𝑖1 + 𝜆2𝑏𝑖2+. . . +𝜆𝑘𝑏𝑖𝑘
equation of the
APT Number of
Risk-free rate Sensitivity of an factors k
𝜆0 asset i to factor j
What are those
Factor betas or
factors?
Factor loadings How many do we
Equation of APT have?
𝐸 𝑟𝑖 = 𝑟𝑓 + [𝐸 𝑟𝑀 − 𝑟𝑓 ]𝛽𝑖 Ross doesn’t say!!
similar to CAPM
 Assumptions and equation
7

 Risk factors ( )
 Factor affecting the return on all assets (common to
all assets)
 These factors must be identified and can be of several
kinds (macroeconomic, microeconomic, statistical,
etc.)
Coefficients are
 Sensitivities to risk factors (bik)
often estimated
 The sensitivity factor () impacts each of the securities
differently using a statistical
technique ‘factor
 Thus, each security "i" will have a particular
sensitivity for each of the risk factors "k“ measured by analysis’
bik

However, when it is time to apply the theory, there is no indication of what these
factors represent… And how many!
 B - Comparing the CAPM and the APT
8

CAPM APT
Form of the equation Linear Linear
Number of risk factors 1 K (1)
Premium of the risk factor E(Rm)-Rf {λi}
Sensitivity to the risk factor i {bik}
« zero-beta » return Rf Rf
 Remarks:
 APT admits a universe where securities are influenced by more than one factor.
 The graphical representation as well as the modeling of the investment universe
are potentially highly complexified for APT.
 APT allows for greater flexibility in the design and risk nature of securities.
 C - Using the APT: Two-stock and a two-factor model
example
9

 Assume that there are two common factors: one related to unexpected changes in
the level of inflation and another related to unanticipated changes in the real level
of GDP.

 Risk factor definitions and sensitivities:

δ1 = The risk premium related to this factor is 2 percent for every 1 percent change in the rate (λ = 0.02).
δ2 = The average risk premium related to this factor is 3 percent for every 1 percent change in the rate growth
(λ2 = 0.03).
λ0 = rate of return on a zero-systematic risk asset (zero-beta) is 4 percent (λ0 = 0.04) = rf

 Assume also that there are two assets (x and y) that have the following sensitivities
to these common risk factors:
bx1 = response of Asset x to changes in the inflation factor is 0.50
bx2 = response of Asset x to changes in the GDP factor is 1.50
by1 = response of Asset y to changes in the inflation factor is 2.00
by2 = response of Asset y to changes in the GDP factor is 1.75
 Using the APT: Two-stock and a two-factor model example
10

 Under the APT model, what should be the expected return of the 2
securities?
𝐸 𝑟𝑖 = 𝑟𝑓 + 𝜆1𝑏𝑖1 + 𝜆2𝑏𝑖2+. . . +𝜆𝑘𝑏𝑖𝑘 ** See Reilly &
𝐸(𝑟𝑖) = 𝑟𝑓 + 𝜆1𝑏𝑖1 + 𝜆2𝑏𝑖2 Brown
Chapter 7
𝐸 𝑅𝑋 = 0.04 + 0.02 ∗ 0.5 + 0.03 ∗ 1.50 = 9.5%
And appendix
𝐸(𝑅𝑌 ) = 0.04 + 0.02 ∗ 2.00 + 0.03 ∗ 1.75 = 13.25%
AOA
If the prices of the two assets do not reflect expected returns, we would expect investors
to enter into arbitrage arrangements selling overpriced assets (short) and using the
proceeds to purchase the underpriced assets until the relevant prices are corrected.

Price relationships that satisfy the no-arbitrage condition are important because we expect
them to hold in real-world markets.
 When it comes to putting theory into practice…
11

 CAPM framework specifies the single risk factor: excess return to

the market portfolio. Process of estimation quite straightforward.
 Challenge: a set of restrictive assumptions and potential correlated residuals.

 APT is a theoretical and elegant model. Free of many restrictive

assumptions of CAPM.
 Challenge: the inability to identify the risk factors.

 Potential solution to APT? Other multifactor models.

 Section 3: Factor models
12

A. Multifactor models of security market returns are divided into three

types
B. Macroeconomic-based risk factor model
C. Statistical factor model
D. Fundamental-based risk factor model
E. Practical uses of multifactor models
 A - Multifactor models of security market returns are
divided into three types
13

1 Macroeconomic factor models Risk factors can be viewed as

macroeconomic in nature
2 Statistical factor models Risk factors are not specified

3 Fundamental factor models Risk factors can also be viewed at a

microeconomic level

 A wide variety of empirical factor specifications have been employed in

practice. (APT is quite theoretical…).
 Each alternative models attempt to identify a set of economic influences.
 Empirical studies are inconclusive on the number of factors to retain. Thus,
there is a low use of statistical factors.
 Economic factors, allowing for an economic interpretation between risk
factors and the performance of securities or portfolios, are favored.
 B - Macroeconomic-based risk factor model
14

 In a multifactor model, the analyst chooses the exact number and identity of risk
factors, while the APT model does not specify either of them and the CAPM
only specifies one.
 Use the observable economic time series as the factors. Examples: inflation, economic growth,
interest rates, FX rates etc.
 Models assume that security returns responds linearly to the macroeconomic shocks.
 Using forecasts of macroeconomic factors to predict security returns.

Fit = Period t return to the jth

designated risk factor

𝑅𝑖𝑡 = 𝑎𝑖 + 𝑏𝑖1 𝐹1𝑡 + 𝑏𝑖2 𝐹2𝑡 + ⋯ + 𝑏𝑖𝐾 𝐹𝐾𝑡 + 𝑒𝑖𝑡

Rit = Security i’s return that can be 𝑒𝑖𝑡 = is the portion of security i
measured as either a nominal return that may not be explained
or excess return to security i by the factor model
 Macroeconomic-based risk factor model
15

CRR model. Developed by Chen, Roll and Ross (1986)

 The macroeconomic variables considered are:

1. Returns of the NYSE index

2. Monthly growth of industrial production
3. Variation of the CPI
4. Difference between expected and actual CPI
5. Unanticipated changes in the credit spread (=Baa-Rf)
6. Unanticipated changes in the spot yield curve (=LT – ST rf)

 Their findings: The key contribution of their research was to

demonstrate that factors beyond the market portfolio could
significantly explain variations in stock returns, thereby supporting
the multifactor approach of APT. However, economic significance of
risk factors changed over time.
 Macroeconomic-based risk factor model
16

BIRR model. Developed by Burmeister, Ibbotson, Roll

and Ross (1994).

 Analyzed the predictive ability of a model based on the following

set of macroeconomic factors:
1. Confidence risk
2. Time horizon risk
3. Inflation risk
4. Business cycle risk
5. Market timing risk
 Macroeconomic-based risk factor model BIRR
17

 Investor confidence
 Risk associated with unanticipated changes in investor confidence in
investing in risky securities (or portfolios)
 Many financial assets have a positive sensitivity to the confidence risk factor.
An unexpected increase in investor confidence leads to an appreciation in the
value of securities and portfolios with a high positive sensitivity to this factor
 Some securities have a negative sensitivity to the risk factor of confidence:
they are so-called "safe haven“.

 Risk relating to the time horizon

 Risk due to unanticipated changes in the willingness of investors to invest for
a long term
 Growth stocks have a stronger (positive) sensitivity vs value stocks
 Sensitivity to this factor can be positive or negative
 Macroeconomic-based risk factor model : BIRR
18

 Risk relating to inflation

 Risk related to unanticipated changes in the rate of inflation
 An increase in inflation brings downward pressure on the stock market
 Securities with the highest sensitivities :
 Securities in the retail sector
 Hotels and leisure services
 Sectors selling luxury goods
 Securities with the lowest sensitivities: consumer staples and other basic
goods and services firms
 Few stocks attract investors in times of high inflation
 Macroeconomic-based risk factor model : BIRR
19

 Risk relating to business cycle

 Risk related to unanticipated changes in the growth rate of the national economic
activity (generally GDP growth rate)
 Sectors performing better in times of economic growth have a positive (high)
sensitivity to this risk factor (e.g. retail)
 stocks in the utilities sector have low and even negative sensitivity to this factor

 Risk relating to market timing

 Linked to unanticipated changes in the stock market that can not be explained by
other macroeconomic factors
 Corresponds to the market risk of the CAPM; interpretation identical to that of
the CAPM beta
 Difference from CAPM: residual market risk measure i.e.. risk not captured by
other factors
 Macroeconomic-based risk factor model : BIRR
20

Risk Factor Risk Premium

If found empirically, the premia are : Confidence 2.59%
Time Horizon -0.66%
Inflation -4.32%
Business Cycle 1.49%
Market timing 3.61%
Source: Rilley and Brown

Example: find the expected return on a market index

Contribution of the risk
Exposure to the risk factor to the expected
Risk Factor Risk Premium of the S&P 500 return of the S&P 500
Confidence 2.59% 0.27 0.70%
Time Horizon -0.66% 0.56 -0.37%
Inflation -4.32% -0.37 1.60%
Business Cycle 1.49% 1.71 2.55%
Market timing 3.61% 1 3.61%
Expected excess return
over the risk-free rate
of the S&P 500 8.09%
 Macroeconomic-based risk factor model : BIRR
21

From the previous results, we can generalize the BIRR model

using the following equation :

𝐸(𝑅𝑖) = 𝑟𝑓 + 𝛽𝑖1(2.59) + 𝛽𝑖2(−0.66) + 𝛽𝑖3(−4.32) + 𝛽𝑖4(1.49) + 𝛽𝑖5(3.61)

where,
Rf : risk-free rate (T-bills)
βij = sensitivity of security i to factor j for j = 1,…5
 C – Statistical factor model
22

 Definition:
 Statistical factor models are models employed to a set of historical returns to
determine factors that explain historical returns
 Objective:
 to look for unrelated (uncorrelated) factors that explain observed historical returns on
securities: APT Tests
 Types:
 Factor analysis models
 Principal component analysis (PCA)

 The actual nature of the factors is not specified by the model. These models make
no assumptions.
 Not very useful for portfolio management: problems of interpretation.
Another drawback is the need for a long and stable history of returns for accurate
estimation of factor betas.
 D - Fundamental-based risk factor model
23

 Characteristic-based approach. Microeconomic models.

 Use observed company attributes as factor betas since they
explain a considerable proportion of common return.
 Examples: market capitalization, P/E ratio, financial leverage etc.
 Factors tend to be correlated for firms in the same industry.
 With this approach, the analyst decides precisely how many and
what attributes need to be estimated

 List of some models:

1) Fama & French three factor model The factors are the
2) Fama & French four factor model returns estimated using
3) And many others… cross-sectional regression
 1) Fama & French three factor model
24

 The FF three-factor model puts three factors forward (1993):

𝑅𝑖𝑡 − 𝑅𝐹𝑅𝑡 = 𝛼𝑖 + 𝑏𝑖1 𝑅M𝑡 − 𝑅𝐹𝑅𝑡 + 𝑏𝑖2 𝑆𝑀𝐵𝑡 + 𝑏𝑖3 𝐻𝑀𝐿𝑡 + 𝑒𝑖𝑡
1. Size of firms. The firm size factor, also known as SMB (small minus big). Equal
to the return to a portfolio of small capitalization stocks less the return to a
portfolio of large capitalization stocks.
2. Book-to-market ratio: HML (High Minus Low): return to a portfolio of stocks
with high ratios of book-to-market values less the return to a portfolio of low book-
to-market value stocks
3. Excess return on the market. The market: market portfolio return minus
return on risk-free rate (market risk premium)

It is a better approach than the Capital Asset Pricing Model (CAPM), as CAPM explains
70% of a portfolio’s diversified returns, whereas Fama-French explains roughly 90%.
 2) Fama & French four factor model
25

 Fama, French & Carhart 1997:

𝑅𝑖𝑡 − 𝑅𝐹𝑅𝑡 = 𝛼𝑖 + 𝑏𝑖1 𝑅M𝑡 − 𝑅𝐹𝑅𝑡 + 𝑏𝑖2𝑆𝑀𝐵𝑡 + 𝑏𝑖3𝐻𝑀𝐿𝑡 + 𝑏𝑖4𝑀𝑂𝑀𝑡 + 𝑒𝑖𝑡

1. SMB & HML & market premium: same definition as before

2. MOM (Momentum): New factor. Return of a portfolio of the best

stocks over the period minus the return of the worst stocks.

The Cahart model is considered a superior one, given its explanatory power of around
95%.
 3) Other models
26

SBB model. Developed by Salomon - Smith - Barney

1. Variation in the anticipated long term economic growth rate (risk of long-term economic
growth): measured by the monthly change in the growth rate of industrial production
2. Short term business cycle (short-term business cycle risk): change in the spread between
the 20-year corporate bond yield and the Treasury bond yield of the same maturity
3. Variation in the yields of long-term bonds (long-term interest rate risk): measured by the
change in the yield on 10-year Treasury bonds
4. Variation in the yields on Treasury bills (short-term interest rate risk): measured by the
change in the yield on short-term treasury bills; usually 1 or 3 months
5. Unanticipated change in the rate of inflation (inflation risk): This is the difference
between the realized inflation rate and the expected inflation rate (by reliable estimation
models!)
6. Change in the value of the dollar against currencies of main trade partners: dollar value
weighted by value of trade with economic partners
7. Beta or residual market risk: residual risk factor not captured by the 6 first factors of the
model
 BARRA model
27

Developed by MSCI Barra 1. Volatility

2. Momentum
 13 fundamental microeconomic
3. Market capitalization
variables are considered. 4. Linearity of earnings vs
market cap.
 Specificities: 5. Volume of transactions
 A factor is specific to the security’s industry 6. Growth
7. Profit Margin
 International specificity 8. Value Index
9. Variability of earnings
10. Financial Leverage
11. Sensitivity to currency
12. Dividend yield
13. Non estimated indicators
 Wilshire/Atlas
28

1. P/E ratio: Last share price divided by the sum of the last 4 quarterly
earnings per share (EPS)
2. Book/Market ratio: Last share price / book value per share
3. Market capitalization
4. Review of net earnings by analysts or measurement of analysts'
momentum : % of analysts who have revised upward earnings per share of
the company; analysts' momentum or enthusiasm about the security
5. “Reversal” : difference between the return realized in the last period on
the security and the "normal" return expected given the beta of the
security
6. The “torpedo” of returns or the measure of profit momentum : measure of
the anticipated growth rate of EPS relative to the historical growth rate of
the company's EPS
7. Historical Beta (CAPM): market (systematic) risk of the security.
Calculated using monthly data for the last 5 years (minimum of 38 latest
monthly observations)
 Goldman Sachs
29

Not marketed (used internally by the company)

1. P/E ratio
2. Retained earnings per share / price
3. Earnings before tax, interest and amortization /(Market value of shares + Book value of long
term debt)
4. Revision of analysts' forecasts over the past 3 months: (Number of upward revision - Number of
downward revisions) / (Total number of analyst forecasts)
5. Momentum of share price : Total return over the last 12 months - Last month's return
6. Sustainable growth : Average (market consensus) long-term growth rate
7. Historical Beta (CAPM) : Estimated over 5 years by regression of excess returns of securities on
excess market return
8. Risk of unpleasant surprise : Risk that the realized EPS are lower than the expected EPS
(similar to torpedo of earnings in the Wilshire Atlas model)
9. Residual Risk : Specific risk, not explained by market variations
 When it comes to putting theory into practice…
30

 It is probably safe to assume that both the CAPM and multifactor

models will continue to be used to value capital assets.

 Further empirical tests of those theories needed.

 The ultimate goal being to determine which theory does the best job
of explaining current returns and predicting future ones.

 Subsequent work in this area will seek to identify the set of factors
that best captures the relevant dimension of investment risk as well as
explore the intertemporal dynamics of the models (for example,
factor betas and risk premia that change over time).
 E– Practical uses of multifactor models
Active management Passive management
31
Risk Management
Models determine the required  These models are used in Exposure of an asset to multiple
rates of return for each security constructing portfolios that risk factors guides decisions about
based on identified risk factors. obtain a consistent result on its inclusion in a portfolio
the characteristics of the according to the risk management
These models help in benchmark. strategy and the level of risk
tolerated by the client
establishing intended
exposures to various risk  Replicates a stock market
Knowledge of the portfolio's
factors. index at a lower cost (reducing
the number of securities to exposure versus that of the market:
reduce transaction costs when allows to identify the manager's
Active managers can thus use bets.
multifactor models to make buying and rebalancing).
bets on desired factors while The presence of securities with
hedging or even remaining different levels of exposure to risk
neutral on the other factors. factors: portfolio diversification
strategy (combining multiple
assets with different risk profiles)
Extract from formula sheet midterm (more formulas
are available for midterm – see sheet ZC/Midterm)
32
Exercices to complete on ZC
33

Q4 & Q5