Multifactor Models and Factor Investing

Vincent Milhau
vincent.milhau@edhec.edu

Academic year 2024-2025

1 / 43
Examples of Multifactor Strategies in Stocks
▶ Blackrock (iShares exchange-traded funds):

▶ Scientific Beta (equity indexes):

2 / 43
Examples of Multifactor Strategies in Stocks (Con’t)
▶ Russell (US equity fund):

3 / 43
Why Are Multifactor Strategies Interesting?
▶ The CAPM’s central prediction, namely

𝜇𝑖 − 𝑅𝑓 = 𝛽𝑖 × [𝜇𝑚 − 𝑅𝑓 ]

(𝜇𝑖 − 𝑅𝑓 = expected excess return on asset 𝑖;

𝛽𝑖 = market beta;
𝜇𝑚 − 𝑅𝑓 = market risk premium)
is not well verified in the data.
▶ Other factors than the market may have explanatory power for
differences between expected returns:
▶ Investors may require a premium for being exposed to other
systematic risk factors than the market (Fama-French model);
▶ Expected returns may depend on other characteristics (e.g.,
industry, country...) than the market beta (Barra risk model).

▶ Knowing asset pricing factors helps us construct efficient

portfolios.

4 / 43
Objectives
▶ Review some classical patterns in stock returns: the size, value
and momentum effects.
▶ Introduce the 3-factor Fama-French model and the 4-factor
extension (Fama-French-Carhart).
▶ Eugene Fama (b. 1939, Nobel Memorial Prize in Economics in 2013)
and Ken French (b. 1954).

▶ Present the principles of factor-based portfolio construction.

5 / 43
Outline

Size, Value and the 3-Factor Model of Fama and French (1993)

Beyond 3 Factors

Factor Investing

6 / 43
Grouping Stocks Into Portfolios
▶ Most of the asset pricing literature has tested the CAPM on stock
portfolios.
▶ Grouping stocks on attributes has a long tradition in empirical
asset pricing:
▶ Black, Jensen and Scholes (1972). “The Capital Asset Pricing Model:
Some Empirical Tests”: group stocks on past estimated beta;
▶ Same in Fama and MacBeth (1973). “Risk, Return, and Equilibrium:
Empirical Tests”;
▶ ... and many papers following Fama and French (1992). “The
Cross-Section of Expected Stock Returns”. Journal of Finance: sort
on size and book-to-market ratio.
▶ Principles:
▶ For every stock, collect an attribute known at the portfolio
formation date: past beta, market cap, book-to-market ratio, past
12-month return...
▶ Make groups of increasing attribute, e.g. 3, 6 or 10 groups;
▶ Calculate the value-weighted or equally weighted returns within
each group;
▶ Hold the stocks until the next portfolio formation date, e.g. 3, 6 or
12 months later.
7 / 43
Early Evidence for the Size Puzzle
▶ Banz (1981). “The relationship between return and market value
of common stocks”. Journal of Financial Economics.
▶ Stocks are sorted on the market capitalization.
▶ Key finding: stocks of large firms earn lower returns than “small”
stocks after controlling for market exposure.

▶ “Arbitrage” portfolios are constructed by taking a long position

in small or average firms and a short position in large or
average firms.
▶ 𝑛 is the numbers of stocks in each leg (long and short), and stocks
are equally weighted within each leg. The market factor is the
excess return on the CRSP value-weighted index.
▶ Alphas are significant, which contradicts the CAPM.
8 / 43
Evidence By Fama and French (1992)
▶ Fama and French (1992). “The Cross-Section of Expected Stock
Returns”. Journal of Finance.
▶ Universe: stocks from NYSE, AMEX and NASDAQ. Sample period:
07.1963 to 12.1990.
▶ Excerpt of their Table II:
▶ Every year, stocks are sorted into 10 groups on their market
capitalization from small (1) to big (10).

▶ The average return decreases from small to big: it is the size

effect.
▶ But the beta decreases too.
9 / 43
Is It Size or Beta?
▶ FF92 also subdivide each size group into sub-groups of 𝛽.
▶ The beta is estimated from the past 24 to 60 months, as available.

▶ Within every beta group, a size effect remains.

▶ The sort on beta creates less dispersion in returns than the sort
on size.
▶ This challenges the CAPM.
10 / 43
Value Investing
▶ Value investing has a long tradition in finance:
▶ Benjamin Graham (1894 - 1976): Security Analysis and The
Intelligent Investor (the latter co-authored with David Dodd (1895 -
1988));
▶ Warren Buffet, CEO of Berkshire Hathaway since 1965.

▶ Key idea: buy stocks that are priced below their “intrinsic value”,
which depends on the company’s ability to make earnings.
▶ And keep a “margin for safety” to account for uncertainty in the
determination of intrinsic value.

▶ Academic version of this approach: a systematic investment

strategy that selects stocks with a low market value relative to
some accounting measure.
▶ The most popular valuation ratio in the academic literature is
the book-to-market ratio:
Book value of equity
𝐵/𝑀 =
Market value of equity

11 / 43
Value Effect
▶ Stocks are now sorted on their book-to-market ratio.

▶ Excerpt of Table IV in FF92:

▶ The average return increases from growth (low B/M) to value

(high B/M) stocks: it is the value effect.
▶ The value effect is a puzzle for the CAPM because it is not
explained by the market beta.

12 / 43
A Long-Short Strategy
▶ On a portfolio formation date, sort stocks into value and growth,
and
▶ Purchase $1 of value stocks;
▶ Sell short $1 of growth stocks;
▶ Invest $1 in the risk-free asset (collateral).

▶ The return on this portfolio is

1 −1 1
𝑟𝑝 = ×𝑟 + × 𝑟growth + × 𝑅𝑓
1 value 1 1
= 𝑅𝑓 + 𝑟value − 𝑟growth

▶ Hence the excess return:

𝑟𝑝 − 𝑅𝑓 = 𝑟value − 𝑟growth

13 / 43
Does the CAPM Explain the Returns on the
Long-Short Strategy?
▶ The workbook Value_effect.xlsx contains the monthly returns on
B/M-sorted and value-weighted portfolios and the market factor
from the data library of Ken French.
▶ We use the same sample period as Fama and French: 1963.07 to
1990.12.
1. Calculate the monthly excess returns on a portfolio that goes long
the 10% most value stocks (“Hi 10”) and short the most growth
ones (“Lo 10”).
2. Calculate the average excess return on the long-short portfolio.
3. Regress the excess returns of the long-short portfolio on the
market factor. Does the factor explain the profitability of the
strategy?

14 / 43
Size and Value Factors
▶ Fama and French (1993). “Common Risk Factors in the Returns on
Stocks and Bonds”. Journal of Financial Economics.
▶ FF93 sort NYSE, AMEX and NASDAQ stocks on market
capitalization and book-to-market ratio:
▶ 2 size groups: Small, Big.
▶ 3 book-to-market groups: Low, Medium, High.

Definitions
The size factor is the excess return of a portfolio of small socks over
a portfolio of big stocks:

SMB = 𝑟small − 𝑟big

The value factor is the excess return of a portfolio of value stocks

over a portfolio of growth stocks:

HML = 𝑟value − 𝑟growth

15 / 43
Factor Construction Details
▶ 6 value-weighted portfolios are formed from the size and
book-to-market groups:
S/L, S/M, S/H, B/L, B/M, B/H
▶ Stocks are value-weighted in each of the 6 portfolios.
▶ Small-minus-big factor:

𝑆/𝐿 + 𝑆/𝑀 + 𝑆/𝐻 𝐵/𝐿 + 𝐵/𝑀 + 𝐵/𝐻

SMB = −
3 3
▶ High-minus-low factor:

𝑆/𝐻 + 𝐵/𝐻 𝑆/𝐿 + 𝐵/𝐿

HML = −
2 2
▶ These factors (and many other datasets) are available from the
data library of Ken French.

16 / 43
Fama-French Alpha and Betas
▶ Inputs:
▶ Returns on security or portfolio 𝑖: 𝑟 .;
𝑖
▶ Market, size and value factors (MKT, SMB, HML);
▶ Time series of risk-free rates: 𝑅 .
𝑓

Fama-French Model
The market, size and value betas in the Fama-French model are the
multivariate betas of excess returns on security or portfolio 𝑖 with
respect to the market, size and value factors:

𝑟𝑖 − 𝑅𝑓 = 𝛼𝑖 + 𝛽𝑖/MKT MKT + 𝛽𝑖/SMB SMB + 𝛽𝑖/HML HML + 𝜀𝑖

The alpha in the Fama-French model is the intercept of the

regression.

▶ Coefficients are estimated by running a multivariate time-series

linear regression.

17 / 43
Return Decomposition in Fama-French Model
▶ Take the expectation and in both sides of the regression
equation:

𝔼 [𝑟𝑖 − 𝑅𝑓 ] = 𝛼𝑖 + 𝛽𝑖/MKT ΛMKT + 𝛽𝑖/SMB ΛSMB + 𝛽𝑖/HML ΛHML

ΛMKT , ΛSMB , ΛHML = factor risk premia

▶ Factor premia are given by

ΛMKT = 𝔼 [𝑟𝑚 − 𝑅𝑓 ]

ΛSMB = 𝔼 [𝑟small − 𝑟big ]

ΛHML = 𝔼 [𝑟value − 𝑟growth ]

▶ The Fama-French factors are said to be pricing factors for asset 𝑖

if 𝛼𝑖 = 0.
▶ The econometrics problem is to test if 𝛼𝑖 = 0 given an estimate
𝛼̂ 𝑖 and a standard error SE [𝛼̂ 𝑖 ].
18 / 43
The Fama-French Model and Asset Pricing
▶ The Fama-French factors are said to be pricing factors for asset 𝑖
if 𝛼𝑖 = 0.
▶ If the FF factors are pricing factors, expected returns are
functions of betas, but they do not depend on the idiosyncratic
2
variance 𝜎𝜀,𝑖 .
▶ Idiosyncratic risk is not rewarded by additional expected return:
it is “bad risk”!
▶ Exposure to pricing factors is rewarded by a higher expected
return: it is “good risk”!
▶ To know if the FF factors are pricing factors, we need to test if
𝛼𝑖 = 0 for a set of test assets.
▶ See your econometrics course for how to calculate the estimate 𝛼̂
𝑖
and the standard error SE [𝛼̂ 𝑖 ] and how to design the t-test of the
hypothesis that 𝛼𝑖 = 0.

19 / 43
One-Factor versus Three-Factor Regression
▶ See workbook Factor_regressions.xlsx.
▶ We have monthly quotes (10.2019 - 07.2024) for the S&P 500
index and the GE Aerospace (General Electric until April 2024)
stock.
▶ Also given are the Fama-French factor values and the monthly
risk-free rate multiplied by 100.
▶ Exercise:
1. Calculate the monthly excess returns multiplied by 100.
2. Regress the two series of excess returns on the market factor
alone, then on the Fama-French factors, and calculate the
t-statistics of the alphas and the betas.
3. Compare the betas with respect to the market factor and the R2
across the two regressions. Why do the betas and the R2 differ
across the two regressions?
4. Calculate the sample averages of the Fama-French factors and the
two series of excess returns.
5. Calculate the Fama-French implied excess return and check that
the sum of this quantity and the alpha equals the sample average
excess return.
20 / 43
𝑅2 Increases as New Factors Are Added
▶ Illustration with 2 factors:

𝑟𝑖 = 𝛼𝑖 + 𝛽1/𝑖 𝐹1 + 𝛽2/𝑖 𝐹2 + 𝜀𝑖

▶ Stack observations in vectors

𝐫
𝐅1 , 𝐅2 and 𝐫.
▶ Linear regression is
𝛆̂ equivalent to the search of
Linear combinations the best approximation of 𝐫
of factors by a linear combination of
the factors.
𝐫̂
▶ The best approximation 𝐫̂ is
𝐅2 𝐅1
the orthogonal projection of
𝐫 on the regressors.
▶ Approximation is more accurate when the number of factors
increases.
▶ Coefficient of determination: 𝑅2 = ‖𝐫‖
̂ 2 /‖𝐫‖2 .
21 / 43
Outline

Size, Value and the 3-Factor Model of Fama and French (1993)

Beyond 3 Factors

Factor Investing

22 / 43
Momentum Effect
▶ Jegadeesh and Titman (1993).“Returns to Buying Winners and
Selling Losers: Implications for Stock Market Efficiency”. Journal
of Finance.
▶ Investment strategy:
▶ Sort stocks on their past 𝐽-month return (𝐽 = 3, 6, 9 or 12);
▶ Group stocks in 10 equally weighted portfolios;
▶ Buy the 10% winners and sell the 10% losers (zero-dollar
strategy);
▶ Hold the portfolio for 𝐾 months;
▶ Go back to step 1.

▶ The excess returns of this long-short strategy are calculated

from 01.1965 to 12.1989.

23 / 43
Momentum Effect in Stock Returns (Con’t)
▶ Excerpt of Table I in JT93 (average monthly excess returns, in %):

Momentum Effect in Stocks

If stocks are sorted on their return over the past 3, 6, 9 or 12 months
and held for another 3, 6, 9 or 12 months, past winners earn higher
average returns than past losers.
24 / 43
Horizon Matters!
▶ Fama and French (1996). “Multifactor Explanations of Asset
Pricing Anomalies”. Journal of Finance.
▶ Excerpt of their Table VI (sample period from 07.1963 to 12.1993; 1
= past losers; 10 = past winners):

▶ Stocks are equally weighted and held for 1 month.

▶ Up to ~24 months, we have a momentum (continuation) effect .

▶ At 60 months, we have a reversal effect .

25 / 43
Do the Fama-French Factors Explain the Momentum
Effect?
▶ Excerpt of Table VII (sample period from 07.1963 to 12.1993) in
Fama and French (1996):

▶ Although the 3 factors explain at least 75% of the time-series

variance, there is a momentum pattern in alphas.
▶ 7 alphas out of 10 are significant at the 5% level (𝑡 < −1.96 or
𝑡 > +1.96).
▶ The Gibbons-Ross-Shanken statistic is “large” (p-value = 0.000),
leading to reject the null that 𝛼1 = 𝛼2 = ⋯ = 𝛼10 .
26 / 43
Long-Term Reversal and the 3-Factor Model
▶ Excerpt of Table VII (sample period from 07.1963 to 12.1993) in
Fama and French (1996):

▶ The 3-factor model explains the average returns on portfolios

sorted on past 5-year return: alphas are insignificant.

27 / 43
Momentum Factor
▶ Introduced by Carhart (1997). “On Persistence in Mutual Fund
Performance”. Journal of Finance.

Definition
The winners-minus-losers (WML) factor, or momentum factor, is the
excess return of a portfolio of past winners over a portfolio of past
losers.
▶ Carhart’s (1997) momentum factor:
▶ Stocks are sorted on past 12-month return, skipping last month,
into 3 groups (30% winners, 30% losers).
▶ Stocks are equally weighted in long and short legs and are held for
1 month.
▶ Momentum factor from Ken French’s data library:
▶ Stocks are sorted in 2 size groups (S, B) and 3 past 12-month return
groups (30% W, 30% L).
▶ They are value-weighted and held for 1 month.
𝑆/𝑊 + 𝐵/𝑊 𝑆/𝐿 + 𝐵/𝐿
Mom = −
2 2
28 / 43
The Fama-French-Carhart Model
Fama-French-Carhart Alpha and Betas
The market, size, value and momentum betas are the multivariate
betas of excess returns on security or portfolio 𝑖 with respect to the
market, size, value and momentum factors:

𝑟𝑖 − 𝑅𝑓 = 𝛼𝑖 + 𝛽𝑖/MKT MKT + 𝛽𝑖/SMB SMB + 𝛽𝑖/HML HML + 𝛽𝑖/WML WML + 𝜀𝑖

29 / 43
What’s Next? The Volatility Puzzle
▶ Ang, Hodrick, Xing and Zhang (2006). “The Cross Section of
Volatility and Expected Returns”. Journal of Finance.
▶ Stocks are sorted every month into 5 groups based on their past
idiosyncratic volatility in the FF93 model (volatility of residuals)
and are value-weighted.
▶ Excerpt of their Table VI (sample period from 07.1963 to 12.2000):

▶ The high volatility-minus-low volatility portfolio earns a

negative return that the FF93 model does not explain.

30 / 43
How Many Factors Do We Need?
▶ Profitability and investment effects in stock returns: Fama and
French (2006), Novy-Marx (2013).
▶ Survey by Harvey, Liu and Zhu (2016) “... and the Cross Section of
Expected Returns”, Review of Financial Studies: more than 300
factors have been proposed!
▶ Widely accepted stock market factors carrying a premium:
▶ Market, size, value (Fama-French 3-factor model);
▶ Momentum (Fama-French-Carhart 4-factor model);
▶ Profitability and investment;
▶ Volatility (more debated).
▶ What in other asset classes (bonds, commodities...)?
▶ Empirical evidence for momentum and “value” (defined as
long-term reversal) effects in bonds, commodities and currency
futures: Asness, Moskowitz and Pedersen (2013). “Value and
Momentum Everywhere”. Journal of Finance.

▶ To validate a factor, we need robust historical evidence and an

economic justification.
31 / 43
Microeconomic Justification for Factors
▶ Foundation paper: Rosenberg and Marathe (1976). “Common
Factors in Security Returns: Microeconomic Determinants and
Macroeconomic Correlates”.
▶ Some macroeconomic, regulatory, geopolitical... events can
impact groups of firms.
▶ Changes in monetary policies impact the financial sector and
leveraged firms;
▶ Oil prices matter for oil companies, the car industry and
energy-intensive industries;
▶ ...

▶ Central assumption: firms with similar characteristics should

have similar risk and similar expected returns.
▶ Industry implementation by MSCI: Barra models for stocks,
bonds (Government and credit), currencies...
▶ Many characteristics of a stock can enter such a model: country,
sector, leverage, profitability...

32 / 43
How Can We Justify the Size, Value and Momentum
Effects?
▶ There are many possible stories!
▶ Concerns about risk:
▶ Small firms and firms with high B/M might be closer to distress:
Fama and French (1992, 1993), Vassalou and Xing (2004).
▶ Small firms are less liquid (Amihud and Mendelson, 1986).
▶ Small and value firms might be more exposed to undiversifiable
risk factors in the sense of the Arbitrage Pricing Theory or the
Intertemporal Capital Asset Pricing Model (see next slides).
▶ Market imperfections and behavioral explanations:
▶ Investors lack information about small firms (Banz, 1981), are
excessively optimistic about growth stocks (Lakonishok et al.,
1994)...
▶ New information is slowly reflected in stock prices, which could
explain momentum (Hong and Stein, 1999).
▶ ...

33 / 43
A Theoretical Justification for Multiple Factors:
Arbitrage Pricing Theory
▶ Foundation paper: Ross (1976). “The arbitrage theory of capital
asset pricing”. Journal of Economic Theory.
▶ If returns have a 𝐾-factor structure
𝑟𝑖̃ = 𝑐𝑖 + 𝛽𝑖/1 𝐹1 + ⋯ + 𝛽𝑖/𝐾 𝐹𝐾 + 𝜂𝑖
and there is no arbitrage in the market, then 𝜇̃ 𝑖 converges to 𝛃𝑖 𝚲
for some vector of factor premia as the specific variance 𝕍 [𝜂𝑖 ]
shrinks to 0.
▶ A well-diversified portfolio is a portfolio with little or no specific
variance.
▶ A formal mathematical definition in a universe with infinitely many
assets is given by Chamberlain (1983). “Funds, Factors and
Diversification in Arbitrage Pricing Models”. Econometrica.
▶ Well-diversified portfolios are still exposed to factors, so factors
are undiversifiable.
▶ Investors require premia for being exposed to “adverse”
undiversifiable factors.
34 / 43
Another Theoretical Justification:
Intertemporal Capital Asset Pricing Model
▶ Foundation paper: Merton (1973). “An Intertemporal Capital
Asset Pricing Model”. Econometrica.
▶ Investors rebalance their portfolios in continuous time and face
a randomly changing short-term interest rate.
▶ Welfare / utility is a function of the horizon, the current wealth
and the current interest rate:
𝐽(𝜏 , 𝑊 , 𝑟)
▶ A 3-fund separation theorem obtains, i.e., the optimal portfolio
of every investor is a combination of:
▶ The market portfolio;
▶ The risk-free asset (cash account);
▶ The interest rate-hedging portfolio, i.e., long-term bonds.
▶ The positive hedging demand for bonds is due to the fact that
bond prices increase (good news) when interest rates decrease
(bad news).
▶ This is true for investors who have more marginal utility from one
additional dollar when interest rates are low (𝐽𝑊𝑟 < 0).
35 / 43
Intertemporal Capital Asset Pricing Model (Con’t)
▶ The additional hedging demand for bonds increases bond
prices, hence decreases the expected returns on bonds at
market equilibrium.
▶ More generally, if investors face time-varying investment
opportunities, they will take:
▶ Long positions in assets that pay off well when investment
opportunities get worse;
▶ Short positions in assets that pay off well when opportunities
improve.

▶ Eventually, the expected return in equilibrium will depend on

how the asset covaries with investment opportunities.
▶ Fama-French factors could also be factors in the sense of the
ICAPM: Petkova (2006). “Do the Fama–French Factors Proxy for
Innovations in Predictive Variables?”. Journal of Finance.

36 / 43
Outline

Size, Value and the 3-Factor Model of Fama and French (1993)

Beyond 3 Factors

Factor Investing

37 / 43
Conceptual Justification
▶ Reminder (2-fund separation theorem): under the CAPM’s
assumptions, a portfolio is efficient if, and only if it is a
combination of the market portfolio and the risk-free asset.
▶ Multifactor extension (𝐾-fund separation theorem): if there are
𝐾 pricing factors (e.g., Fama-French or Fama-French-Carhart)
and these factors are the excess returns on factor-replicating
portfolios, then a portfolio is efficient if, and only if, it is a
combination of the 𝐾 factor-replicating portfolios and the
risk-free asset with factor betas proportional to

𝛀−1 𝚲

𝛀 = factor covariance matrix;

𝚲 = vector of factor premia
▶ Formal statement of this result: Proposition 1 of Grinblatt and
Titman (1987). “The Relation Between Mean-Variance Efficiency
and Arbitrage Pricing”. Journal of Business.

38 / 43
Mathematical Proof
▶ Denote with 𝐖 the matrix of percentage weights of the 𝐾
factor-replicating portfolios (𝑁 × 𝐾).
▶ The factor premia are the expected excess returns on the
portfolios, hence
𝚲 = 𝐖′ 𝛍̃
▶ Because the factors are pricing factors, we have

𝛍̃ = 𝛃𝚲

where 𝛃 is the 𝑁 × 𝐾 matrix of constituents’ betas.

▶ The factor covariance matrix is

𝛀 = 𝐖′ 𝚺𝐖

▶ The multivariate betas are given by

𝛃 = 𝚺𝐖𝛀−1

39 / 43
Mathematical Proof (Con’t)
▶ If a portfolio is efficient, then, by the 2-fund separation theorem,
the risky asset weights satisfy

𝐰 = 𝑥𝐰tan

where 𝑥 = 𝜎/𝜎
̄ tan and 𝜎̄ is the target volatility.
▶ The factor exposures of the portfolio are

𝛃𝑝 = 𝛃 ′ 𝐰
= 𝑥𝛀−1 𝐖′ 𝚺𝐰tan
𝑥
= 𝛀−1 𝐖′ 𝛍̃
𝑒
𝑥 −1
= 𝛀 𝚲
𝑒

where 𝑒 = 1/ [𝟏′ 𝚺−1 𝛍].

̃

40 / 43
Mathematical Proof (Con’t)
▶ Reciprocal: Suppose now that the exposures of a portfolio 𝐰
have the form
𝛃𝑝 = 𝑥𝛀−1 𝚲
for some 𝑥.
▶ Then,
𝛀−1 𝐖′ 𝚺𝐰 = 𝑥𝛀−1 𝚲
▶ Hence,
𝐖′ [𝚺𝐰 − 𝑥 𝛍]̃ = 𝟎
▶ Hence,
𝐰 = 𝑥𝚺−1 𝛍̃ + 𝚺−1 𝐡
where 𝐖′ 𝐡 = 𝟎.
▶ We have

𝛍̃ = 𝛃𝚲
= 𝚺𝐖𝛀−1 𝚲

41 / 43
Mathematical Proof (Con’t)
▶ Therefore, if 𝐰 is in the column space of 𝐖, then

𝚺−1 𝐡 = 𝐰 − 𝑥𝐖𝛀−1 𝚲

is in the column space of 𝚺−1 𝐡 too.

▶ Rewrite
𝚺−1 𝐡 = 𝐖𝐮
▶ Then,

𝐖′ 𝐡 = 𝐖′ 𝚺𝐖𝐮
= 𝛀𝐮
=𝟎

▶ But 𝛀 is non-singular, so 𝐮 = 𝟎, hence 𝐡 = 𝟎, and

𝐰 = 𝑥𝚺−1 𝛍̃

is the risky weight vector of an efficient portfolio.

42 / 43
Factor Investing Principles
▶ Efficient portfolios are portfolios of factors, with no idiosyncratic
risk and efficient factor exposures.
▶ Therefore, efficient portfolio construction requires:
▶ Eliminating idiosyncratic risk, which carries no premium.
▶ Controlling / managing exposures to pricing factors.

▶ Efficient portfolio construction is a form of risk management!

▶ In a Homework problem, you will simulate the returns of a
simple equally weighted multifactor portfolio (market, size,
value).

▶ More advanced forms of multifactor

allocation are introduced by Amenc
et al. (2014). “Risk Allocation, Factor
Investing and Smart Beta:
Reconciling Innovations in Equity
Portfolio Construction”. EDHEC-Risk
Institute Publication.
43 / 43