CS计算机代考程序代写 scheme finance ER AI Excel KE1068

KE1068

May 31, 2018

©2018 by the Kellogg School of Management at Northwestern University. !is case was prepared by Professor Phillip
A. Braun. Cases are developed solely as the basis for class discussion. Cases are not intended to serve as endorsements,
sources of primary data, or illustrations of e”ective or ine”ective management. Some details may have been #ctionalized
for pedagogical purposes. To order copies or request permission to reproduce materials, call 800-545-7685 (or
617-783-7600 outside the United States or Canada) or e-mail .edu. No part of this publication
may be reproduced, stored in a retrieval system, used in a spreadsheet, or transmitted in any form or by any means—
electronic, mechanical, photocopying, recording, or otherwise—without the permission of Kellogg Case Publishing.

P H I L L I P A . B R A U N

Smart Beta Exchange-Traded
Funds and Factor Investing

It was early 2015 and executives in the iShares Factor Strategies Group were considering the
launch of some new exchange-traded funds (ETFs). !e new ETFs were in a class of ETFs called
smart beta funds. Most traditional ETFs’ portfolio weights were based on the market capitalization
of a stock (stock price times the number of outstanding shares), but smart beta ETFs’ weighting
schemes were based on #rms’ #nancial characteristics or properties of their stock returns.

iShares was a division of BlackRock, Inc., an international investment management company
based in New York. In 2014, BlackRock was the world’s leading asset manager, with over $4.5
trillion in assets under management.1 In 2014, iShares globally o”ered over 700 ETFs with almost
$800 billion in net asset value in the US and $1 trillion globally—a 39% market share, making
iShares the largest issuer of ETFs in the world.2

!e new smart beta multifactor ETFs being considered by iShares would provide investors
with simultaneous exposure to four fundamental factors that had shown themselves historically
to be signi#cant in driving stock returns: the stock market value of a #rm, the relative value of a
#rm’s #nancial position, the quality of a #rm’s #nancial position, and the momentum a #rm’s stock
price has had. While each of these factors existed in di”erent combinations and di”erent forms
in ETFs already in the marketplace, no #rm was currently o”ering these four as a combination
in a multifactor ETF. !e executives at iShares were unsure whether there would be demand in
the marketplace for such multifactor ETFs, since their value added from an investor’s portfolio
perspective was unknown.

This document is authorized for use only in Prof. Fong wai ming’s FINA6112 FC & FD Term 1, 22 Aug 2021-22, Prof. Fong & Sam (2) at The Chinese University of Hong Kong (CUHK) from Aug
2021 to Feb 2022.

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N G KE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

Factors
Smart beta portfolios were driven by academic research that shows there is a common set of

driving forces, or factors, that consistently explains stocks’ average returns and systematic risks.
Professors Eugene Fama and Kenneth French have been at the forefront of developing factor models,
having written a series of papers examining di”erent #nancially based factor models of stock returns.
As the culmination of their decades of research, in 2013 Fama and French introduced a model that
showed a systematic relationship between the average returns of stocks and #ve underlying factors.3

Based on the capital asset pricing model (CAPM) theory, the #rst factor Fama and French
considered was the relative covariance of a #rm’s stock returns with a market portfolio. !e CAPM
theory states that stock return movements can be broken down into two components: movements
due to the returns of an underlying market portfolio (a market factor) and #rm-speci#c movements,
with a stock’s average returns determined by its co-movements with the market portfolio. !is can
be seen via the CAPM’s factor model:

( ), , ,i t f i i m t f i tr r r r− = + − + (1)
where ri,t is the return at time t for some security i, rf is the return to a risk-free asset, αi and βi are
regression coe$cients, rm,t is the return to a market portfolio (the market factor) at time t, and
εi,t is the regression’s residual. A stock’s co-movement with the market portfolio is measured by its
CAPM beta (βi), the regression coe$cient above.

Figure 1 shows the average returns on #ve portfolios ranked by the size of their betas using
annual data from 1964 through 2014. !ese portfolios were created by sorting all stocks listed
on the NYSE, AMEX, and NASDAQ by their betas and then splitting the stocks into quintile
portfolios based on the level of their beta, with those stocks with betas in the bottom 20% of
the distribution put into a smallest beta quintile portfolio, then the next 20% into the next
highest quintile portfolio, and so on. Figure 1 shows no evidence of a strong upward sloping
relationship between a portfolio’s beta and its average return, contrary to the prediction of the
CAPM theory. !us, Fama and French concluded that the CAPM theory does not explain average
returns very well.*4

!e second factor Fama and French explored was the size of a #rm as measured by its market
capitalization, what is termed a size factor. Fama and French (and others) found a signi#cant
negative relationship between the size of a #rm and its average return—that is, #rms with small
market capitalizations earn higher average returns across time than #rms with large market
capitalizations. Figure 2 plots the average returns of portfolios grouped into quintiles by their
market capitalizations. What we see in Figure 2 is the small #rm e”ect; small #rms have higher
average returns than large #rms, with a consistent rise in average returns as we go from the smallest
to the largest #rm’s portfolio.

* Note that Fama and French used a longer time frame, from 1927–1990, to construct the portfolios in this study
than were used to construct the portfolios in Figure 1 (1964–2014).

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N GKE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

Figure 1: Relationship Between Average Returns and the CAPM Beta

10%

12%

14%

Smallest Betas Quintile 2 Quintile 3 Quintile 4 Largest Betas

A
ve

ra
ge

R
et

ur
n

Source: Author’s calculations using data from Ken French’s data library, accessed August 2016, http://mba.tuck.dartmouth.edu/
pages/faculty/ken.french/data_library.html. These portfolios are from French’s univariate portfolios, sorted by beta, using annual
value-weighted data from 1964–2014.

Figure 2: Relationship Between Average Returns and Firm Size

10%

12%

14%

16%

18%

20%

Smallest Firms Quintile 2 Quintile 3 Quintile 4 Biggest Firms

A
ve

ra
ge

R
et

ur
n

Source: Author’s calculations using data from Ken French’s data library, accessed August 2016, http://mba.tuck.dartmouth.edu/
pages/faculty/ken.french/data_library.html. These portfolios are from French’s univariate portfolios, sorted by firm size (market
capitalization), using annual equal-weighted data from 1964–2014.

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N G KE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

At #rst glance it is not obvious why smaller #rms should have higher average returns. Fama and
French showed that, controlling for other factors, size is related to a #rm’s pro#tability; small #rms
have lower earnings on assets than big #rms.5 Firm size is also thought to proxy for speci#c risk
factors associated with smaller #rms; researchers have explored the underlying sources of such risks,
but the results are not conclusive. Some have argued that because small #rms’ stocks do not trade
as often as larger #rms and are thus less liquid, investors in smaller #rms require higher returns for
accepting this liquidity risk.6 Others suggest that size is correlated with information uncertainty—
that is, smaller #rms are not often followed by investment banks while simultaneously having more
volatile fundamentals.7

!e third factor that Fama and French considered was a value factor, which is the ratio of a
#rm’s book value (as given by a #rm’s balance sheet, its assets minus its liabilities) to its market
capitalization—its book-to-market (B/M) ratio.8 Figure 3 plots the average returns of #ve
portfolios that were created by ranking stocks by their B/M ratio and then sorting them into
quintile portfolios from low B/M ratios to high. What we see in Figure 3 is that the portfolios with
higher B/M ratios have higher average returns, which is termed a value e”ect.

Figure 3: Relationship Between Average Returns and the Book-to-Market Ratio

10%

12%

14%

16%

18%

Lowest B/M Quintile 2 Quintile 3 Quintile 4 Highest B/M

A
ve

ra
ge

R
et

ur
n

Source: Author’s calculations using data from Ken French’s data library, accessed August 2016, http://mba.tuck.dartmouth.edu/
pages/faculty/ken.french/data_library.html. These portfolios are from French’s univariate portfolios, sorted by the B/M ratio, using
annual value-weighted data from 1964–2014.

Fama and French argue that this positive relationship between average returns and B/M ratios
is because high B/M stocks are less pro#table and are relatively distressed, so these #rms are riskier
and have higher average returns to re%ect that. Conversely, low B/M #rms have high returns on
capital with sustained pro#tability (hence they are termed growth stocks); therefore, they are less
risky and have lower average returns. !e B/M e”ect is called a value e”ect because “When a #rm’s

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N GKE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

market value is low relative to its book value, then a stock purchaser acquires a relatively large
quantity of book assets for each dollar spent on the #rm. When a #rm’s market price is high relative
to its book value the opposite is true.”9

!e fourth factor that Fama and French proposed was a pro#tability factor, the ratio of a
#rm’s operating pro#t to its book value.* In Figure 4, the average returns of #ve portfolios are
presented, sorted from low to high pro#tability. To create these portfolios, #rms were ranked by the
Fama and French pro#tability measure and then sorted into quintile portfolios. As can be seen in
Figure 4, more pro#table #rms earn signi#cantly higher average returns than less pro#table #rms.
!e argument is simply that #rms with productive assets should have higher returns than #rms
with unproductive assets. !is pro#tability factor is sometimes referred to as a quality factor—
#rms with higher pro#tability are higher quality #rms. “While traditional value strategies #nance
the acquisition of inexpensive assets by selling expensive assets, [a quality strategy] exploits a
di”erent dimension of value, #nancing the acquisition of [quality] productive assets by selling
unproductive assets.”10

Figure 4: Relationship Between Average Returns and Profitability

10%

12%

14%

Lowest
Profitability

Quintile 2 Quintile 3 Quintile 4 Highest
Profitability

A
ve

ra
ge

R
et

ur
n

Source: Author’s calculations using data from Ken French’s data library, accessed August 2016, http://mba.tuck.dartmouth.edu/
pages/faculty/ken.french/data_library.html. These portfolios are from French’s univariate portfolios, sorted by profitability, using
annual value-weighted data from 1964–2014.

!e #fth Fama and French factor was an investment factor, which they measured as the
percentage change in the value of a #rm’s assets over the course of a year.11 In Figure 5, #rms are
ranked by this investment measure and then sorted into quintile portfolios, from conservative (low

* Fama and French’s measure of operating pro#t is a #rm’s revenues minus costs of goods sold, minus selling,
general, and administrative expenses, minus interest expense.

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N G KE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

levels of investment) to aggressive (high levels of investment). As shown in Figure 5, portfolios with
conservative levels of investment have higher average returns than aggressive #rms.

Fama and French’s rationale for this investment e”ect is that, holding a #rm’s expected
revenues constant, a rise in a #rm’s investment implies lower future expected earnings and thus
lower expected returns, while lower investment levels yield higher expected earnings and thus
higher expected returns. Like the pro#tability factor, the investment factor is also termed a quality
factor because #rms with lower levels of investment are higher quality #rms because they are not
depressing their earnings via excessive investment and are putting less stress on their balance sheets.

Figure 5: Relationship Between Average Returns and Investment

10%

12%

14%

16%

Lowest
Investment

Quintile 2 Quintile 3 Quintile 4 Highest
Investment

A
ve

ra
ge

R
et

ur
n

Source: Author’s calculations using data from Ken French’s data library, accessed August 2016, http://mba.tuck.dartmouth.edu/
pages/faculty/ken.french/data_library.html. These portfolios are from French’s univariate portfolios, sorted by investment, using
annual value-weighted data from 1964–2014.

Separately from Fama and French, a series of authors have examined what is termed a
momentum factor.12 !e momentum e”ect documents that stocks that have large price appreciation
in one year continue to have high price appreciation the following year and stocks with negative
or low price appreciation continue to do so the following year. In Figure 6, stocks are sorted
into quintile portfolios based on their stock price appreciation in the previous year, from those
stocks that performed in the bottom 20% of all stocks in the previous year, what is termed a loser
portfolio, to stocks that performed in the top 20% over the last year, a portfolio of winners. As
Figure 6 shows, the greater the price appreciation for a portfolio over the last year, the higher that
portfolio’s average return.

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N GKE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

Figure 6: Relationship Between Average Returns and Momentum

10%

15%

20%

25%

Lowest
Momentum

Quintile 2 Quintile 3 Quintile 4 Highest
Momentum

A
ve

ra
ge

R
et

ur
n

Source: Author’s calculations using data from Ken French’s data library, accessed August 2016, http://mba.tuck.dartmouth.edu/
pages/faculty/ken.french/data_library.html. These portfolios are from French’s univariate portfolios, sorted by momentum, using
annual equal-weighted data from 1964–2014.

In a recent study, Fama and French found that, after controlling for the market, size, value,
pro#tability, and investment factors, a momentum factor had no signi#cance in determining
average returns.13 Others, however, #nd their evidence inconclusive, and continue to see a relevance
for a momentum factor even after these other factors are controlled for.14

!ere are two schools of thought as to why we see a relationship between a stock or portfolio’s
momentum and its average returns: the rational “markets are e$cient” school and the behaviorist
“markets are ine$cient” school. A rational market perspective on the causes of the momentum e”ect
would hold that #rms with high momentum face greater cash %ow risks and/or higher discount
rates because of the nature of their investment sets.15 !e behaviorists view the momentum e”ect
as resulting from investors’ cognitive biases. For example, investors who choose to invest in a
high-return stock are simply extrapolating past performance into the future, exhibiting a kind
of herding behavior.16

To understand how factors are a set of driving forces across #nancial securities’ returns, the
next step is to consider an actual factor model. A factor model is a regression that relates securities’
returns to a set of factor portfolios. For example, the CAPM theory implies that there is only one
factor driving security returns, the market portfolio. Given this, equation (1) is the CAPM’s factor
model. For the six factors just discussed, the factor model would be a six-factor regression model
of the form:

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N G KE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

( ), , , / ,

, , , ,

( ) ( )

( ) ( ) ( )

i t f i i m t f i Size t i B M t

i Profitabilty t i Investment t i Momentum t i t

r r a b r r s F h F

r F c F m F

− = + − + +

+ + + +
(2)

where ri,t is again the return on some stock, rm,t is the return on the market, rf is the return to a
risk-free asset, FSize,t through FMomentum,t are the factor portfolios for each of the other #ve variables
discussed above, ai is the regression intercept, bi to mi are the regression slope coe$cients, and εi,t is
the regression’s error term. From this regression you can see that this six-factor model is the CAPM
shown in equation (1) plus #ve additional factors.

!e factor portfolios, FSize,t through FMomentum,t, can be created in di”erent ways. A simpli#ed
example of one standard approach is to use the decile portfolios with the highest average return
for a particular variable—for example, the decile with the smallest #rms would be used for the
size factor portfolio, FSize,t; and the decile with the highest B/M ratios would be used for the B/M
or value factor portfolio, FB/M,t, and so on. !ese factor portfolios are called long-only portfolios
because they only include purchases of stocks in the portfolio.

An example of a second standard approach, following Fama and French’s work, is to use a
long-short factor portfolio. Simplistically, a long-short factor portfolio is constructed by taking the
returns on the decile with the highest average return less the returns on the decile with the lowest
average return. It is called long-short because the portfolio includes both purchases (long) as well
as shorts—you short a security by borrowing the security from a broker and then selling it. For
example, using this approach the size factor portfolio, FSize,t, would be the returns to the decile
with the smallest #rms less the returns to the decile portfolio with the biggest #rms.* Fama and
French call this the SMB portfolio, for small #rms minus big #rms. Fama and French’s other long-
short portfolios for their #ve-factor model are:

• HML (high minus low): !e returns of the decile portfolio that includes the #rms with
the highest B/M ratios less the returns to the decile portfolio with the lowest B/M ratios.

• RMW (robust minus weak): !e returns of the decile portfolio that includes the #rms
with the most robust pro#tability less the returns to the decile portfolio with the
weakest pro#tability.

• CMA (conservative minus aggressive): !e returns of the decile portfolio that includes the
#rms with the lowest (most conservative) levels of investment spending less the returns to
the decile portfolio with the highest (most aggressive) investment spending.

Background on Smart Beta ETFs
!e intent behind smart beta ETFs is to capture the high expected returns identi#ed for factors

in the work of Fama and French and others.17 !e term “smart beta” is a fairly new marketing

* !is simpli#es what Fama and French actually do, which is to take the bottom 30% of #rms in market size for
the small #rm portfolio and the top 30% of #rms in market size for the big #rm portfolio. !is is also true for the
other Fama-French long-short portfolios de#ned in the case.

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N GKE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

identi#er; some of the alternative names that have been used include strategic beta, active beta,
enhanced index, alternative beta, and scienti#c beta. !ere is no one speci#c de#nition of what a
smart beta portfolio is. However, a common theme across de#nitions is that smart beta portfolios
are constructed such that the emphasis is weighting stocks in these portfolios not on the traditional
measure of market capitalization, but by incorporating into their weighting scheme some aspect
of a security’s fundamental value, such as a stock’s B/M ratio, pro#tability, or a characteristic of
the security’s performance, such as a stock’s momentum.* Regardless what de#nition is used for
smart beta ETFs, the bottom line is that they are proxies for the factor portfolios, the Fi,t’s, that we
discussed in the previous section.

Smart beta ETFs are considered a combination of passive and active investing. !e funds are
passive because they passively mimic what are termed factor indexes and hence do not require any
input from a portfolio manager. !ey are considered active because their weights deviate from
standard market capitalization weights. As Cli” Asness, a founder, managing principal, and chief
investment o$cer at AQR Capital Management, a global investment management #rm that has
been at the forefront in creating momentum factor portfolios, has stated, “A portfolio that deviates
from market weights .&.&. must be balanced by other investors who are willing to take the other
side of those bets. For example, for every value investor, who tilts toward or selects cheap value
stocks, there must be an investor on the other side who is underweighting value and overweighting
expensive, growth stocks. Hence, as everything must add up to the market-weighted portfolio,
everyone at once cannot hold or tilt toward value at the same time.”18 Smart beta portfolios are
not active in the sense that fund managers are searching for mispriced securities. Although some
justify smart beta portfolios from the behaviorist perspective that the factors are mispriced,19 it is
not necessary to use this justi#cation to motivate smart beta strategies.20

Besides smart beta ETFs that capture the size, value, quality, and momentum factors we have
discussed, there are also dividend, minimum-volatility, and other factor-based smart beta stock
ETFs, as well as bond ETFs that capture factors speci#c to bonds. Dividend ETFs in general try
to capture potentially higher average returns from investing in stocks paying high dividends.†21
!e rationale for minimum-volatility (also called low volatility) ETFs is the documented evidence
that stocks with low volatility or risk have higher average returns than high volatility stocks.22
!ere is also a class of multifactor ETFs, whose goal is to provide exposure to a set of two or more
factors simultaneously.

In 2014, mutual funds that mimic underlying factors had existed for a while, but smart beta
ETFs were newer to the marketplace.‡ Figure 7 shows the level of assets under management in

* !is is not a strict de#nition, however; some #rms classify traditional value, growth, dividend, and small #rm
portfolios whose weighting scheme uses market capitalization weights into the smart beta category, while other
#rms do not.

† Note that Fama and French consider the B/M ratio and their pro#tability measure to be better predictors of
average returns than the dividend yield; see https://famafrench.dimensional.com/questions-answers/qa-dividends-
is-bigger-better.aspx, accessed January 2018.

‡ !e market statistics in this section are the author’s calculations using data from the Center for Research in
Security Prices and a partial list of ETFs from http://www.etf.com, accessed September 2016. Note that the group
of smart beta ETFs included in this analysis excludes traditional value, growth, dividend, and small #rm portfolios
whose weights are based on market capitalizations, which ETF.com classi#es as smart beta.

1 0

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N G KE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

smart beta ETFs listed in the US from 2003, when the #rst smart beta ETFs were introduced,
through 2014. !e smart beta market grew from around $1.5 billion in 2003 to just over $160
billion at the end of 2014, a cumulative annual growth rate of 47%.* Figure 7 also shows the
breakdown of assets under management by smart beta category, with dividend ETFs having the
most assets under management and momentum ETFs with the least. Figure 8 shows the number
of smart beta ETFs and their share of the overall US ETF market. In 2003, there were only six
smart beta ETFs; this had grown to 277 by 2014. In 2003, smart beta ETFs were 1% of the overall
US ETF market and this grew to an 8% market share by 2014.

Figure 7: Market Size and Breakdown for US Smart Beta Exchange-Traded Funds

$20

$40

$60

$80

$100

$120

$140

$160

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

N
et

A
ss

et
V

al
ue

(B
ill

io
ns

$
) Momentum

Growth
Quality
Other
Bonds
Volatility
Size
Multifactor
Dividend
Value

Sources: Author’s calculations using data from the Center for Research in Security Prices and a partial list of smart beta ETFs from
http://www.etf.com. Note that the group of smart beta ETFs in the analysis excludes traditional value, growth, dividend, and small
firm portfolios whose weights are based on market capitalizations.

* If traditional market capitalization–based value, growth, and small #rm ETFs are included in the analysis, the
smart beta market size in 2014 was $360 billion.

1 1

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N GKE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

Figure 8: Number of US Smart Beta Exchange-Traded Funds and Market Share
Relative to Total US ETF Market

100

150

200

250

300

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

N
um

be
r

of
F

un
ds

P
er

ce
nt

ag
e

M
ar

ke
t S

ha
re

Number of Smart Beta Funds Smart Beta Share of Total ETF Market

Sources: Author’s calculations using data from the Center for Research in Security Prices and the list of ETFs from
http://www.etf.com. See the source notes to Figure 7 for more detail.

In 2014, iShares had a 38% overall US ETF market share and a 22% share of the smart beta
market. Its closest competitor, State Street Corporation, had a 23% overall ETF market share, but
only a 10% smart beta market share. !e other major player in the smart beta market was Invesco
PowerShares Capital Management, which had a 15% smart beta market share, but only a 5%
overall ETF market share. !e Vanguard Group had a 21% overall ETF market share but o”ered
no smart beta ETFs.*

At the end of 2014, iShares o”ered 19 di”erent smart beta ETFs in the US with $36 billion in
assets under management. A list of these ETFs is presented in Table 1, along with their net asset
values. iShares introduced its #rst dividend ETF in 2003, the iShares Select Dividend ETF (DVY),
and by 2014, iShares o”ered a range of smart beta ETFs, both international and domestic US, for
equities as well as bonds. In 2013, iShares introduced four smart beta ETFs directly related to the
research of Fama and French and others: its size, value, quality, and momentum factor ETFs.

* Vanguard did not sell smart beta funds because it questioned the value the approach added for investors. See, for
example, C.&B. Philips et al., “An Evaluation of Smart Beta and Other Rules-Based Active Strategies,” Vanguard
Research (2015). Vanguard did, however, o”er a variety of capitalization-weighted ETFs similar to smart beta
ETFs, such as its dividend appreciation ETF (VIG), its value ETF (VTV), and its small cap ETF (VB).

1 2

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N G KE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

Table 1: List of Smart Beta ETFs Sold by iShares in 2014

Name Ticker
Net Asset Value

($ in millions)

iShares Asia/Pacific Dividend ETF DVYA 55

iShares Core Dividend Growth ETF DGRO 152

iShares Core High Dividend ETF HDV 5,197

iShares Edge MSCI Min Vol Asia ex Japan ETF AXJV 5

iShares Edge MSCI Min Vol EAFE ETF EFAV 1,356

iShares Edge MSCI Min Vol Emerging Markets ETF EEMV 1,907

iShares Edge MSCI Min Vol Europe ETF EUMV 4

iShares Edge MSCI Min Vol Global ETF ACWV 1,561

iShares Edge MSCI Min Vol Japan ETF JPMV 10

iShares Edge MSCI Min Vol USA ETF USMV 3,581

iShares Edge MSCI USA Momentum Factor ETF MTUM 482

iShares Edge MSCI USA Quality Factor ETF QUAL 721

iShares Edge MSCI USA Size Factor ETF SIZE 213

iShares Edge MSCI USA Value Factor ETF VLUE 517

iShares Emerging Markets Dividend ETF DVYE 213

iShares International Select Dividend ETF IDV 4,163

iShares MSCI USA Equal Weighted ETF EUSA 62

iShares Select Dividend ETF DVY 15,554

iShares Yield Optimized Bond ETF BYLD 9

Sources: Center for Research in Security Prices and http://www.etf.com. Note that traditional market capitalization–weighted
portfolios are excluded from this analysis.

!e multifactor ETFs that iShares was considering would be a combination of the size, value,
quality, and momentum factors together, encompassing both international and domestic US
stocks. As shown in Figure 7, multifactor ETFs only had $40 million in net asset value in 2003,
which grew to $18 billion by 2014, a 76% cumulative annual growth rate. Prior to 2013, all the
multifactor ETFs in the market captured just two e”ects, combining the size e”ect with one of
the other factors. !e type of multifactor ETFs that iShares was considering, which capture three
or four factors simultaneously, was very new to the market. At the end of 2014, only three #rms
o”ered such multifactor products in the US market, with a combined net asset value of only about
$150 million.*23

* A small competitor, WisdomTree Investments Inc., introduced a multifactor ETF based on the size, quality, and
dividend factors in 2013, the WisdomTree US Small Cap Quality Dividend Growth Fund (DGRS), which at the
end of 2014 only had a market value of $25 million. In mid-2014, State Street o”ered a set of 12 international
multifactor ETFs, its SPDR MSCI Quality Mix ETF suite, and had announced the release of a US multifactor
ETF for early 2015. !ese MSCI Quality Mix ETFs combined the value, low volatility, and quality factors into
one ETF; at the end of 2014, this group of ETFs only had a total market value of $66 million. J.&P. Morgan Asset
Management introduced two multifactor ETFs in 2014: the JPMorgan Diversi#ed Return Global Equity ETF
(JPGE) and the Diversi#ed Return International Equity ETF (JPIN), which brought together the size, value,
momentum, and volatility factors and combined had around $60 million market value at the end of 2014.

1 3

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N GKE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

Smart Beta Construction Methodology
A critical aspect of smart beta portfolios is that the portfolio construction methodology is

rule-based, transparent, and low-cost to implement,24 thus ensuring that the portfolios #t within
the standard structure of passive ETFs. No common construction methodology is used across the
smart beta ETFs sold by di”erent #rms—each uses di”erent #nancial variables and methods of
calculating weight. Because of this, the performance of di”erent smart beta ETFs with the same
objective can diverge.

Perhaps the easiest way to understand the essence of how smart beta portfolios are created is
to consider the smart beta index construction methodology pioneered by Research A$liates, an
institutional investment advisor and manager and a pioneer in the development of smart beta
indexes and portfolios, called fundamental indexing.25 Given a #nancial variable, such as book
value, the fundamental indexing approach determines the weight for each stock in an index via the
formula:

,
,

,
1

i j
i j N

i j
i

F
w

F
=

=
(3)

where wi,j is the fundamental weight for some security i for some #nancial variable j, Fi,j is the
value of the #nancial variable j (such as book value) for #rm i and N is the number of securities in
the sample.

Often smart beta ETFs use more than one #nancial variable in their construction. For example,
Research A$liates uses sales, cash %ow, dividends, and book value to create its RAFI Value Factor
US Index.26 To accomplish this, the weights for each #nancial variable j from above are averaged
across all the #nancial variables used to de#ne the index and then re-standardized.

!e methodology that iShares used in constructing its size, value, quality, and momentum
factor ETFs was developed by MSCI Inc., a New York–based #nancial services #rm known for
its indexes, performance reporting, and risk management tools. !e following iShares smart beta
ETFs were indexed to the following MSCI factor indexes:

• !e iShares Edge MSCI USA Size Factor ETF (SIZE) was indexed to the MSCI
USA Risk Weighted Index,* in which the weights were determined by the risk of the
underlying stocks.27

• !e iShares Edge MSCI USA Value Factor ETF (VLUE) was indexed to the MSCI
USA Enhanced Value Index, in which the weights were calculated using three valuation
characteristics: the forward price-to-earnings ratio, the enterprise value to operating cash
%ow ratio, and the price-to-book ratio (the inverse of B/M).28

* Note that the iShares size factor ETF is not a traditional small #rm portfolio because the weights in the portfolio
are determined by the risk of a stock, not its market capitalization.

1 4

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N G KE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

• !e iShares Edge MSCI USA Quality Factor ETF (QUAL) was indexed to the MSCI
USA Quality Index, which used the variables return on equity, debt to equity, and earnings
variability to determine its stock weightings.29

• !e iShares Edge MSCI USA Momentum Factor ETF (MTUM) was indexed to the MSCI
USA Momentum Index, which used a stock’s returns for the previous 6 to 12 months,
standardized by the stock’s volatility, to estimate the stock weightings.30

In constructing their factor indexes, MSCI assigns each stock in their sample what is called a
Z-score. !e Z-score is a statistic based on where a #rm’s valuation characteristic (e.g., its price-
earnings ratio) lies relative to the mean of the valuation characteristic for the sample of stocks used,
standardized by the standard deviation of the valuation characteristic across the sample. Speci#cally,

,
,

( )
( )

i j j
i j

F F
z

F
−

= (4)

where zi,j is the Z-score, Fi,j the #nancial variable j under consideration for #rm i, µ(Fj ) is the
mean of #nancial variable j for all stocks in the sample, and σ(Fj ) is the standard deviation of that
#nancial variable across all stocks in the sample. Z-scores have an advantage over fundamental
indexing because they consider not just the mean of the underlying #nancial variable, but also the
dispersion of the underlying #nancial variable. You can interpret a Z-score as a measure of how
many standard deviations a #nancial variable is from its mean.

MSCI also uses multiple #nancial variables in de#ning its indexes. As mentioned above,
for example, its MSCI USA Enhanced Value Index uses the forward price-to-earnings ratio,
the enterprise value to operating cash %ow ratio, and the price-to-book ratio. To get the weight
for a security in its Enhanced Value Index, an individual Z-score is calculated for each of the
three #nancial variables used to construct the index and then a composite Z-score is calculated
as the average of the individual Z-scores across the #nancial variables j from above, and
then re-standardized.31

!e multifactor ETFs that iShares was considering would be constructed using the methodology
of MSCI’s diversi#ed multifactor indexes. MSCI produced these indexes for international markets
as well as the US, for both large-mid-cap stocks and small-cap stocks. MSCI planned to introduce
these indexes in early 201532 and iShares was considering o”ering ETFs to coincide with their
introduction. Corresponding to the MSCI indexes, iShares was considering introducing #ve
multifactor ETFs: a US large-mid-cap, a US small-cap, an international large-mid-cap (excluding
US), a global large-mid-cap (including US), and an international small-cap. In creating its
diversi#ed multifactor indexes, MSCI assigned a weight to each stock that was an average of the
stock’s individual Z-scores for the size, value, quality, and momentum indexes.*

* MSCI’s size weighting scheme for its multifactor index was di”erent than that used to create its risk weighted
index, which was used by iShares to create its size ETF. In the multifactor indexes a stock is assigned a size weight
based on the inverse of the log of a stock’s market capitalization (rather than a stock’s risk, as in the risk weighted
index). Note that once each stock is assigned an overall Z-score, the index is optimized to match the standard
deviation of MSCI’s USA Index.

1 5

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N GKE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

MSCI’s multifactor indexes were constructed di”erently than the multifactor ETFs already
in the marketplace. !ese other multifactor ETFs were constructed by averaging the underlying
single-factor indexes, a top-down approach. MSCI used a bottom-up approach, weighting each
stock in the multifactor index by the set of its fundamental valuation characteristics, the Z-scores.
Alain Dubois, head of new business and product development in the index unit at MSCI, claimed
that the bottom-up approach taken by MSCI’s indexes “will optimize the exposure to these factors
to a greater degree than [the top-down approach].”33

Factor Investing
Everyone has heard that they should hold a well-diversi#ed portfolio, but the question is, how

should one diversify one’s portfolio? A standard method is to diversify investments by asset class—
that is, investing some in equities, some in bonds, some in real estate, some in commodities, and
so on. Factor investing o”ers an alternative method to implement a diversi#cation strategy. With
factor investing you diversify your portfolio across factor portfolios (smart beta ETFs) so that your
investments will have some amount in a small #rm portfolio, some in a value portfolio, some in a
quality portfolio, and so on. As described in iShares promotional material, “Smart beta strategies
are the ‘gateway’ to factor investing—designed to harvest broad, persistent drivers of returns by
taking advantage of economic insights, diversi#cation, and e$cient trading execution.”34

!e intuition behind factor investing is that a standard diversi#cation strategy that emphasizes
asset classes may not have the optimal exposures to the factors that are known to yield high average
returns across time. Moreover, since diversi#cation is also about the correlation structure across
investments, factor investing can also help maximize diversi#cation by providing a richer set of
cross-correlations and thereby better manage the risk and return of a portfolio across market cycles.35
!is is because the returns and risks between factor portfolios vary relative to each other across
changing business conditions while traditional asset class market capitalization stock portfolios’
returns and risks move more closely together. For example, the returns and risks of size and value
factors are highest in the early part of an economic expansion, while momentum and quality factors
are highest at the beginning of an economic contraction.36 Furthermore, quality and momentum
factors outperform in declining interest rate environments, while value and size outperform with
rising rates.37 !ese #ndings indicate that the return and risk of a factor investment portfolio
are managed by bringing the di”erent cyclicality of the factors together, mitigating the e”ect of
changing business conditions.

Given that there is strong cyclicality in factor returns, with certain factors underperforming
for signi#cant periods of time and outperforming in others, the question arises as to whether an
investor can tactically move between the di”erent factor portfolios in anticipation of changing
economic conditions to get a higher return. !is process of switching investments between di”erent
factors across time is called style or factor timing.

Academic and industry experts have debated the ability of investors to know when to switch
between factor portfolios and the bene#ts of switching. A series of authors have claimed that they
are able to predict factor movements ex ante and therefore are able to construct tactical portfolios

1 6

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N G KE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

that outperform a buy-and-hold strategy.38 But other authors have examined the performance of
mutual funds that switch between factors and found that their performance is worse than funds
that maintain a consistent exposure to all of the factors.39 Cli” Asness of AQR Capital Management
found “timing strategies to be quite weak historically.”40 Rather than trying to time factors, he
recommends that investors “instead focus on identifying factors that they believe in over the very
long haul, and aggressively diversify across them.”41 Robert Arnott, the founder and chairman of
Research A$liates, and his co-authors take the alternative perspective that “active timing of smart
beta strategies and/or factor tilts can bene#t investors. We #nd that performance can easily be
improved by&emphasizing&the factors or strategies that are trading cheap relative to their historical
norms and by&deemphasizing the more expensive factors or strategies.”42

Data Analysis
In its decision process about introducing multifactor ETFs, iShares wanted to consider the

value added of its potential US large-mid-cap multifactor ETF relative to the size, value, quality,
and momentum factor ETFs it already sold, and how it compared to the traditional investing
approach using asset classes (e.g. large-cap, mid-cap, international, etc.), which were market
capitalization–based portfolios.

To put its analysis in perspective, iShares wanted to reexamine Fama and French’s #ve-factor
analysis and the momentum factor. !e original Fama and French study used data from 1964–
2012 but iShares wanted to know if those #ndings held over shorter periods of time. iShares was
particularly interested in 2005–2014, which was the period for which it had data for its smart beta
factor ETFs. !e data for this analysis is in Exhibit 1.

To understand the bene#ts of factor investing relative to traditional asset class–based
investing, iShares compared its potential US multifactor ETF relative to other ETFs it already
sold. !e monthly returns from 2005–2014 for di”erent ETFs or their proxies are presented
in Exhibit 2.* !e Markowitz minimum-variance frontiers for di”erent scenarios are presented
in Exhibits 3 and&4.

Conclusion
Although there was substantial evidence that the returns for individual factor portfolios

outperformed traditional market capitalization portfolios, there was no such evidence for
multifactor portfolios, so their value added from an investment perspective was still an open
question. !erefore, it was still uncertain whether there would be a need or a demand in the
marketplace for multifactor ETFs. iShares was not only the market leader in the overall ETF
market, but also the leader in the smart beta ETF market. To avoid tarnishing its brand name,
iShares would not introduce multifactor ETFs unless their value was proven.

* !e MSCI factor indexes’ returns are used as proxies because the actual iShares ETFs are new to the marketplace
and do not have a long time series of returns. When MSCI created its factor indexes it provided a history of the
data, so the time series available to analyze is long enough to study.

1 7

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N GKE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

Since iShares already sold smart beta ETFs that targeted individual factors, it had to compare
the performance of the new multifactor ETFs to its existing smart beta ETFs, as well as to traditional
market capitalization ETFs. !e executives in the Factor Strategies Group still had some important
analysis to conduct before they made their #nal decision about these new multifactor ETFs.

1 8

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N G KE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

Exhibit 1: Historical Data for Fama and French’s Financial Characteristics
Decile Sorted Portfolios

See the Exhibits Excel file
Source: See the source notes to Figures 1–6 for sources, with the difference being that this is monthly data for the period 2005–
2014 and the data used in those figures is annual data for 1964–2014.

Exhibit 2: Monthly Returns for Selected MSCI Indexes and iShares ETFs

See the Exhibits Excel file
Source: The data for the MSCI indexes is from http://www.MSCI.com, accessed August 2016, and the iShares data is from the
Center for Research in Security Prices.

Exhibit 3: Minimum-Variance Frontiers Calculated from Different
Combinations of the MSCI Factor Indexes and iShares ETFs

See the Exhibits Excel file
Source: Author’s calculations using the data from Exhibit 2.

Exhibit 4: Minimum-Variance Frontiers Calculated from Different
Combinations of the MSCI Factor Indexes and iShares Bond ETF

See the Exhibits Excel File
Source: Author’s calculations using the data from Exhibit 2.

1 9

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N GKE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

Endnotes
1 BlackRock website, accessed September 2016, http://www.blackrock.com.
2 iShares website, accessed September 2016, http://www.ishares.com, and the author’s calculations using data

from the Center for Research in Security Prices.
3 Fama and French’s #rst circulated draft was in 2013; their #nal published paper is: E. F. Fama and K. R. French,

“A Five-Factor Asset Pricing Model,” Journal of Financial Economics 116 (2015): 1–22.
4 E. F. Fama and K. R. French, “!e Cross-Section of Expected Stock Returns,” Journal of Finance 47 (1992):

441–65.
5 E. F. Fama and K. R. French, “Multifactor Explanations of Asset Pricing Anomalies,” Journal of Finance 51

(1996): 55–84.
6 See M. Bennan, T. Chordia, and A. Subrahmanyam, “Alternative Factor Speci#cations, Security Characteristics,

and the Cross-Section of Expected Stock Returns,” Journal of Finance 49 (1998): 345–73.
7 See X. F. Zhang, “Information Uncertainty and Stock Returns,” Journal of Finance 61 (2006): 105–37.
8 See, for example, S. Basu, “!e Relationship Between Earnings Yield, Market Value, and Returns for NYSE

Common Stocks: Further Evidence,” Journal of Financial Economics 12 (1983): 129–56.
9 R. Novy-Marx, “!e Other Side of Value: !e Gross Pro#tability Premium,” Journal of Financial Economics 108

(2013): 1–2.
10 Ibid., 1.
11 For background, see S. Titman, K. Wei, and F. Xie, “Capital Investments and Stock Returns,” Journal of Financial

and Quantitative Analysis 39 (2004): 677–700. Also see G. Aharoni, B. Grundy, and Q. Zeng, “Stock Returns
and the Miller Modigliani Valuation Formula: Revisiting the Fama French Analysis,” Journal of Financial
Economics 110 (2013): 347–57.

12 For background of momentum, see N. Jegadeesh and S. Titman, “Returns to Buying Winners and Selling
Losers: Implications for Stock Market E$ciency,” Journal of Finance 48 (1993): 65–91, and M. M. Carhart, “On
Persistence in Mutual Fund Performance,” Journal of Finance 52 (1997): 57–82. Also see C. S. Asness, “Variables
!at Explain Stock Returns,” PhD dissertation, University of Chicago, Graduate School of Business (1994).

13 E. F. Fama and K. R. French, “Dissecting Anomalies with a Five-Factor Model,” Review of Financial Studies 29
(2016): 1–35.

14 See, for example, C. S. Asness et al., “Fact, Fiction, and Value Investing,” Journal of Portfolio Management 42
(2015): 34–52.

15 C. S. Asness, T. J. Maskowitz, and L. H. Pedersen, “Value and Momentum Everywhere,” Journal of Finance 48
(2013): 929–85.

16 D. K. Hirshleifer and A. Subrahmanyam, “Investor Psychology and Security Market Under- and Over-
Reactions,” Journal of Finance 53 (1998): 1839–85.

17 For details on what ETFs are, see J. Hill, D. Nadig, and M. Hougan, “A Comprehensive Guide to Exchange-
Traded Funds (ETFs),” CFA Research Foundation, May 2015.

18 C. S. Asness et al., “Fact, Fiction, and Value Investing,” Journal of Portfolio Management 42 (2015): 34–52.
19 See, for example, R. Arnott et al., “How Can ‘Smart Beta’ Go Horribly Wrong?” Research A$liates, February

2016, and R. Arnott, N. Beck, and V. Kalesnik, “Timing ‘Smart Beta’ Strategies? Of Course! Buy Low, Sell
High!” Research A$liates, September 2016.

20 See C. S. Asness, “How Can a Strategy Still Work If Everyone Knows About It?” AQR, 2015.
21 See, for example, R. Ball, “Anomalies in Relationsips Between Securities’ Yields and Yield Surrogates,” Journal

of Financial Economics 6 (1978): 103–26.
22 See, for example, R. A. Haugen and N. L. Baker, “Commonality in the Determinants of Expected Stock

Returns,” Journal of Financial Economics 41 (1996): 401–39. More recently an alternative perspective of the
volatility e”ect as a CAPM beta e”ect has been studied; see A. Frazzini and L. Pedersen, “Betting Against Beta,”
Journal of Financial Economics 111 (2014): 1–25.

2 0

S M A R T B E T A E T F S A N D F A C T O R I N V E S T I N G KE1068

K E L L O G G S C H O O L O F M A N A G E M E N T

23 H. Bell, “Daily ETF Watch: SSgA’s US Quality Mix SPDR,” December 22, 2014, http://www.etf.com/sections/
daily-etf-watch/24115-daily-etf-watch-ssgas-us-quality-mix-spdr.html, and JPMorgan Diversi#ed Return Global
Equity ETF, accessed September 2016, https://am.jpmorgan.com/us/en/asset-management/gim/adv/products/d/
jpmorgan-diversi#ed-return-global-equity-etf-etf-shares-46641q100.

24 BlackRock, “Smart Beta Guide,” pg. 12, accessed August 2016, https://www.blackrock.com/au/intermediaries/
literature/whitepaper/ blackrock-smart-beta-guide-en-au.pdf.

25 R. D. Arnott, J. C. Hsu, and P. Moore, “Fundamental Indexation,” Financial Analysts Journal 61 (2005): 83–99.
26 Research A$liates website, accessed September 2016, http://www.researcha$liates.com.
27 MSCI, “MSCI Risk Weighted Indexes Methodology,” June 2014.
28 MSCI, “MSCI Enhanced Value Indexes Methodology,” May 2015.
29 MSCI, “MSCI Quality Indexes Methodology,” August 2014.
30 MSCI, “MSCI Momentum Indexes Methodology,” September 2014.
31 See, for example, “MSCI Enhanced Value Indexes Methodology.”
32 MSCI website, accessed September 2016, http://www.msci.com.
33 L. Woodall, “Index Providers Clash Over Evolution of Multi-Factor Products,” October 2, 2015,

http://www.risk.net/structured-products/feature/2427846/index-providers-clash-over-evolution-of-multi-
factor-products.

34 BlackRock, “Factor Investing: Unlocking the Real Drivers of Returns,” accessed August 2016,
https://www.blackrock.com/es/ literature/brochure/factor-investing-unlocking-drivers-of-return-en-lai.pdf.

35 See A. Ang, W. N. Goeztman, and S. Schaefer, “Evaluation of Active Management of the Norwegian Government
Pension Fund—Global, Report to the Norwegian Ministry of Finance,” December 14, 2009.

36 iShares, “Factor !ese Funds into Your Game Plan,” accessed August 2016, https://www.ishares.com/us/
literature/product-brief/ishares-msci-factor-etfs-product-brief.pdf, and Q.& J. Zhang et al., “!e Link Between
Macro-Economic Factors and Style Returns,” Journal of Asset Management 10 (2009): 338–55.

37 J. Stoneberg and B. Smith, “Getting Smart about Beta,” Invesco PowerShares white paper,
accessed August 2016, https://www.invesco.com/static/us/#nancial-professional/contentdetail?
contentId=634″6163339f410VgnVCM100000c2f1bf0aRCRD.

38 See, for example, K. L. Miller et al., “Size Rotation in the U.S. Equity Market,” Journal of Portfolio Management
39 (2013): 116–27.

39 A. J. Corbett, “Are Style Rotating Funds Successful at Style Timing? Evidence from the US Equity Mutual Fund
Market,” working paper, February 2, 2016, https://ssrn.com/abstract=2736672.

40 C. S. Asness, “Resisting the Siren Song of Factor Timing,” AQR, April 15, 2016.
41 Ibid.
42 Arnott, Beck, and Kalesnik, “Timing ‘Smart Beta’ Strategies? Of Course!”

Related Posts