MASTER’S THESIS I 2017

Smart Beta Factor Investing

Enhancing risk-adjusted returns with exposure to risk factors and empirical evidence of market inefficiencies

Alex Mikaelsson and Martin Nilsson

Lund University School of Economics and Management

Department of Economics
Supervisor: Thomas Fischer

 Acknowledgements
First and foremost, we would like to extend our gratitude to Granit Fonder AB, who provided us
with insight into the asset management industry, assistance in the data collection process as well
as valuable advice and guidance during the course of this thesis. We would also like to
acknowledge our supervisor Thomas Fischer, who provided us with much appreciated feedback
throughout the process.

 Abstract

In an attempt to bridge the gap between active and passive investing, Smart Beta strategies have
become a popular alternative for investors given their systematic, rules-based approach to
portfolio construction and historical tendency to capture market inefficiencies. In this thesis, we
examine the performance of Smart Beta strategies versus the S&P 500 and the Euro Stoxx 600
index for time periods 1994-2016 and 2002-2016 respectively. The strategies analyzed are Value,
Size, Sharpe-Momentum, Quality and Low Volatility. Given that factor investing and various
rules-based strategies have previously been studied in academia, we fill the gap in the literature by
providing our own variables to each factor as well as testing their performance across two
geographical regions. The empirical analysis conducted in this thesis indicates that nine out of ten
Smart Beta portfolios outperform their respective benchmark index on a risk-adjusted basis. We
therefore conclude that Smart Beta strategies can serve as a superior alternative to passively
investing in a cap-weighted index, which questions if markets are truly efficient from an asset
allocation standpoint.

Table of Contents
1. Introduction ................................................................................................................ 6
1.1 Background ................................................................................................................................... 6
1.2 Thesis objective and disposition ................................................................................................... 7
2. Theory ........................................................................................................................ 8
2.1 Efficient Market Hypothesis (EMH) ............................................................................................... 8
2.2 Capital Asset Pricing Model (CAPM) ............................................................................................. 9
2.3 Arbitrage Pricing Theory (APT) .................................................................................................... 10
3. Previous studies ........................................................................................................ 12
3.1 Factor investing ........................................................................................................................... 12
3.2 Smart Beta .................................................................................................................................. 13
4. Factors ...................................................................................................................... 15
4.1 Market ......................................................................................................................................... 15
4.2 Value ........................................................................................................................................... 15
4.2.1 Definition of the Value Smart Beta Factor ........................................................................... 16
4.3 Size .............................................................................................................................................. 17
4.3.1 Definition of the Size smart beta factor ............................................................................... 17
4.4 Sharpe-Momentum ..................................................................................................................... 17
4.4.1 Definition of the Sharpe-Momentum Smart Beta factor ..................................................... 18
4.5 Low Volatility .............................................................................................................................. 18
4.5.1 Definition of Low Volatility Smart Beta factor ..................................................................... 19
4.6 Quality ......................................................................................................................................... 19
4.6.1 Definition of the Quality Smart Beta factor ......................................................................... 20
5. Thesis data and methodology ................................................................................... 22
5.1 Thesis data .................................................................................................................................. 22
5.2 Portfolio generation .................................................................................................................... 22
5.2.1 Scoring ................................................................................................................................. 22
5.2.2 Weighting ............................................................................................................................ 24
5.3 Performance analysis .................................................................................................................. 25
5.3.1 Portfolio Return ................................................................................................................... 25
5.3.2 Daily Return Annualized Volatility ....................................................................................... 26
5.3.3 Sharpe-ratio (SR) .................................................................................................................. 26
5.3.4 Excess returns ...................................................................................................................... 26
6. Empirical analysis ..................................................................................................... 27
6.1 Summary results ......................................................................................................................... 27
6.2 Individual Portfolio results .......................................................................................................... 30
6.2.1 Value .................................................................................................................................... 30
6.2.2 Momentum .......................................................................................................................... 32
6.2.3 Low Volatility ....................................................................................................................... 34
6.2.4 Size ....................................................................................................................................... 35
6.2.5 Quality ................................................................................................................................. 37
6.2.6 Portfolio correlation ............................................................................................................ 39
7. Conclusion ................................................................................................................ 40
7.1 Suggested further research ......................................................................................................... 41
References ....................................................................................................................... 42
Appendix A ....................................................................................................................... 45

 Standardized score ............................................................................................................................ 45
The UCITS framework ....................................................................................................................... 46
The Winsor method .......................................................................................................................... 47
Appendix B ....................................................................................................................... 49
Value variables .................................................................................................................................. 49
Quality Variables ............................................................................................................................... 49
Appendix C ....................................................................................................................... 51
Annual winners and losers ................................................................................................................ 51
Appendix D ...................................................................................................................... 57
Value ................................................................................................................................................. 57
Momentum ....................................................................................................................................... 58
Low Volatility .................................................................................................................................... 59
Size .................................................................................................................................................... 60
Quality ............................................................................................................................................... 61
Appendix E ....................................................................................................................... 62
Risk free rate of return ...................................................................................................................... 62

1. Introduction
1.1 Background
The concept of utility is generally referred to as an abstract measurement of total satisfaction an
individual receives from consuming a good or service. According to the neoclassical economic
theory of utility optimization, rational individuals should then be willing to exchange these goods
and services on the market in order to maximize their respective utility. From an investor
perspective, utility is easier to define as rational individuals should behave in a risk-averse manner
in order to achieve the highest amount of expected future return given the lowest amount of
possible risk (Markowitz, 1952). This brings us to the more practical issue that investors face
today, which is creating a portfolio of assets that generate returns in excess of the market.
However, according to neoclassical theory, this should not be possible as long as markets are
efficient (Fama, 1991). Financial markets should in theory be one of the most efficient markets
today given how fast information is transferred and the speed at which prices adapt. The average
time an investor held a security on the New York Stock Exchange in the 1960’s was
approximately eight years (NYSE, 2010) and with the introduction of automated trading and the
increase in the amount of available information, average holding times have been reduced to
weeks or days rather than years. The markets have arguably never been as quick in pricing as
today, leaving little room for the everyday investor to compete for the ever-sought excess returns
on the market. Nevertheless, investors always face the question of whether to trade on the
market with their own knowledge in order to gain additional utility, known as active investing, or
simply follow the market as a whole by investing in the market portfolio, more commonly known
as passive investing (Bowen et. al., 1993). The question of which strategy is the best is still up for
debate so in an attempt to bridge the gap between these two contrasting strategies, Smart Beta
investing has become a popular alternative. Many investors believe Smart Beta strategies combine
the best of both worlds, providing an alternative that expands the opportunities of portfolio
construction. To quote research affiliates, who are renowned for being one of the global leaders
in Smart Beta and asset allocation: “Smart beta strategies are designed to add value by systematically
selecting, weighting, and rebalancing portfolio holdings on the basis of factors or characteristics other than market
capitalization” (RA, 2017).
Despite the flashy name, Smart Beta is not a revolutionary strategy given that related
concepts such as factor investing and rules-based strategies have been around in academia for
decades. In the 1960’s, Jack Treynor (1961), William Sharpe (1964), John Lintner (1965) and Jan
Mossin (1966) proposed the Capital Asset Pricing model (CAPM), which states that the return of

an investment is a function of its exposure to the market factor, Beta. Expanding on this model,
Stephen Ross (1976) proposed the Arbitrage Pricing Theory (APT), which allowed for several
factors to explain asset returns. Since then, several studies have been conducted showing that
exposure to factors such as Size, Value and Momentum have provided returns in excess of the
market. The goal of Smart Beta is therefore to provide investors with exposure to factors that
persistently drive returns, using a transparent and rules-based approach (Amenc et. al., 2015).
One of the first questions that come to mind when discussing the topic of Smart Beta is
whether these strategies constitute an active or passive investment strategy. Smart Beta resembles
passive investing in the implementation process, which as mentioned earlier, is systematic, rules-
based and transparent. At the same time, it also resembles a form of active investing considering
that the objective of Smart Beta is to increase risk-adjusted returns by providing exposure to
certain factors. The diplomatic answer to the above question is that the Smart Beta is somewhere
in between.

1.2 Thesis objective and disposition

The focus of this thesis is to identify whether Smart Beta investing can serve as a superior
alternative to passively investing in a cap-weighted market portfolio. In order to test this, we will
generate Smart Beta portfolios with exposure to risk factors widely acknowledged in academia
such as Value, Size, Quality, Momentum and Low Volatility. We then measure the performance
of these portfolios against traditional cap-weighted stock indices over time. We provide more
width to the empirical analysis by examining the performance of our Smart Beta portfolios
against both a U.S and European benchmark index. We find that nine out of ten portfolios
exhibit superior risk-adjusted returns compared to their respective benchmark indices between
1994 and 2016 in the U.S. and between 2002 and 2016 in Europe.
The paper is divided into seven sections beginning with the introduction, which provides
background on the subject of financial economics and Smart Beta strategies. Section two
provides insight into previous studies relating to Smart Beta. Section three is dedicated to a more
in-depth review of the theory analyzed in this thesis and their relation to Smart Beta. The fourth
section offers a review of the factors analyzed in this thesis along with the variables underlying
each factor. Section five explains the thesis data and methodology with insights into the security
selection, weighting scheme and performance measurements. Section six presents the empirical
results for our study followed by an analysis relating to previous studies. The last section is
devoted to our own conclusions and suggestions to further research.

2. Theory
This section covers established theories within the field of Financial Economics on which Smart
Beta investing has its roots. They are presented to provide background on contemporary asset
pricing theories and serve as the basis for one of the thesis hypotheses this paper delves into: Are
markets truly efficient? In this section, we cover the Efficient Market Hypothesis (EMH), the
Capital Asset Pricing Model (CAPM) and the Arbitrage Pricing Theory (APT).

2.1 Efficient Market Hypothesis (EMH)

The concept of efficient markets was first introduced by Fama et. al. (1969) and theorized around
how prices of securities and other assets adjust when new information is released to markets. An
efficient market was originally defined as one that “...adjusts rapidly to new information.”, but was
revised two decades later as data transmission had become visibly seamless. This allowed Fama
(1991) to redefine the hypothesis of prices to “...fully reflect all available information”. The newer
definition therefore fully disregards cases of asymmetric information in what contemporary
financial economists refer to as efficient markets. In this sense, the hypothesis predicts that it is
impossible to achieve excess returns through stock market expertise, since all available
information is reflected in the price of securities. Furthermore, since all information to date is
reflected in the price of a security, the only factor that can influence prices is tomorrow’s news.
Assuming that news are unpredictable, the hypothesis tells us that the market completely follows
the idea of a random walk in the sense that all subsequent price changes will be a random step
away from the last registered price of the security. This makes outperformance a game of chance
and beating the market consistently is therefore an impossible feat according to the EMH,
rendering smart beta investing a futile operation.
The EMH has been criticized over time where several contemporary financial economists
point to irrational human behavior as one of the main antagonists of the theory. Robert J. Schiller
(2000) discussed that investors tend to exhibit irrational exuberance, which refers to a situation
where people develop misplaced confidence in the economy and financial markets. Periods of
irrational exuberance are characterized by economic agents becoming less risk-averse, inflated
asset prices, increased borrowing and herd behavior. Historical examples include the most recent
financial crisis of 2007-2009, the dotcom bubble of the early 2000s and the period which led up
to the great depression in 1929. Grossman and Stiglitz (1980) also posited that if financial
markets are efficient, and if all information needed is reflected in the price of a security, then
there would be no incentive for investors to spend money on research, which they obviously do.
In this thesis, we attempt to outperform the market index to see if the EMH holds.

2.2 Capital Asset Pricing Model (CAPM)

In the 1960’s Jack Treynor (1961), William Sharpe (1964), John Lintner (1965) and Jan Mossin
(1966) attacked the problem of optimal portfolio selection and introduced us to the perspective
of the CAPM, which is defined as follows:

𝑟" = 𝑟% + 𝛽" (𝑟) − 𝑟% ) (1)

Where 𝑟" is the expected return of asset a, 𝑟% the risk-free rate of return, 𝛽" the systematic risk
factor Beta and 𝑟) the expected return of the market. Stemming from Markowitz’s (1952)
portfolio theory, the CAPM provides a simple one-factor asset pricing model that attempts to
capture excess market return. The theory argues that the expected return of any given security
depends on its exposure the systematic risk factor. Subsequently, the only way an investor can
earn higher return is to take on higher levels of risk. In other words, CAPM essentially captures
the amount of risk premium, (𝑟) − 𝑟% ), investors demand in order to take on a riskier asset.

Figure 1: The security market line (Sharpe, 1999)

In technical terms, the model uses a setting with risk-averse investors who may invest in or
borrow at the risk-free rate*, 𝑟% , and are able to short any assets. Allowing for investments in the
risk-free rate enables investors to secure any portion of its portfolio with a fixed return and thus
marks a minimum level of return for other riskier assets. A security must therefore exhibit returns
in excess of the risk-free rate for it to be an attractive investment. The risk level is captured by
the asset’s beta, 𝛽" , where a higher beta represents a higher return at the cost of more risk.

*
Risk-free rate is typically derived from the 3-month U.S. T-bill. (Appendix E)
9



The CAPM creates a linear security market line (SML), which shows the expected return
contingent on the level of risk associated with each asset (Figure 1). A beta of one signals an
expected return in line with the market and should any firm deviate from the SML, there will
occur profitable opportunities due to arbitrage. The relationship between risk and return should
cause the risk-adjusted return, or Sharpe ratio, to remain constant regardless of portfolio strategy.
Commonly iterated points of criticism towards the CAPM are its reliance on unrealistic
assumptions, such as the assumption that investors have direct access to credit at risk-free rates.
In fact, borrowing for an investor is likely more expensive than what the model suggests. Since
most investors will not be able to borrow at the risk-free rate, the CAPM should therefore
overestimate the expected return of an asset. Empirical evidence suggests that this is true such as
Banz (1981), who claimed that the CAPM overestimates the expected return of large companies
while underestimating the expected return of small ones. Mistrust against the CAPM coupled
with the hunt for returns in excess of the market has investors searching for exposure to
alternative risk factors other than the market. This inevitably has a connection to Smart Beta.

2.3 Arbitrage Pricing Theory (APT)

In 1976, Stephen Ross extended the CAPM framework by proposing the Arbitrage Pricing
Theory (APT). This model builds on the CAPM’s ability to price assets but instead incorporates
several factors that explain returns other than exposure to the market factor. Ross suggested that
there an infinite number of factors that have the ability to explain the return of an asset. The
expected return of a portfolio according to APT is therefore captured by the following equation:

𝑟" = 𝑟% + 𝛽, 𝑓, + 𝛽. 𝑓. + ⋯ + 𝛽0 𝑓0 (2)

Where the rate of return, 𝑟" , depends on the risk-free rate, 𝑟% , the sensitivity of asset 𝑖’s return to
a factor, 𝛽2 , and the risk premium related to that factor, 𝑓2 . This captures the idea that variables
influence the return of an asset in two steps. First, each specific influence (inflation,
unemployment, firm-specific etc.) is determined. Second, the sensitivity to each specific influence
is considered.
By allowing factors to be fluent and interchangeable, Ross’ approach paved the way for
economists to adopt multi-factor thinking into contemporary asset pricing models. The wide
approach allowed the model to be more flexible in its assumptions and therefore applicable in a
wider range of scenarios than its predecessor, the CAPM. However, what the model gains in
customization it loses in utilization. The flexibility and namely the feature to capture any asset-

10



influencers requires rigorous research, and in extreme cases it is difficult to determine enough
explanatory factors for it to have practical application. The idea of multi-factors naturally reached
Smart Beta investing as well, and this paper leans on much of Ross’s ideas as we use multiple
underlying variables to create our factors.

11



3. Previous studies
In this section, we attempt to provide general results from previous, well-known studies relevant
to this thesis. We start by examining famous papers within factor investing followed by more
recent studies regarding Smart Beta.

3.1 Factor investing

Two professors and colleagues at the University of Chicago, Eugene Fama and Kenneth French
(1993) established what constitutes one of the cornerstones in contemporary factor investing,
namely the Fama and French three-factor model. This model expands on the CAPM by adding
factors Value and Size. They proposed a regression model as follows:

𝑟 − 𝑅% = 𝑅% + 𝐵5 𝑥 𝑅) − 𝑅% + 𝑏8 𝑥 𝑆𝑀𝐵 + 𝑏; 𝑥 𝐻𝑀𝐿 + 𝛼 (3)

Where r is the return of the portfolio, 𝑅% is the risk-free rate of return and 𝑅) the return of the
market portfolio. SMB stands for “small minus big” and states that by buying small cap stocks
and shorting large cap stocks, investors can earn returns in excess of the market over the long
run. HML stands for “high minus low” and essentially states that by buying Value stocks, i.e.,
stocks with low price-to-fundamental ratios, and shorting growth stocks, investors can also earn
returns in excess of the market. Fama and French eventually discovered that a portfolios beta,
i.e., exposure to the market factor, explained around 70% of excess returns. By adding the size
and value factors to the CAPM, the model’s explanatory power jumps to 95%. Put simply, small-
cap and value stocks consistently outperform the market on a regular basis. Expanding on the
three-factor model was Carhart (1997), who proposed a four-factor model to include a
Momentum factor:

𝑟 − 𝑅% = 𝑅% + 𝐵5 𝑥 𝑅) − 𝑅% + 𝑏8 𝑥 𝑆𝑀𝐵 + 𝑏; 𝑥 𝐻𝑀𝐿 + 𝑏) 𝑥 𝑊𝑀𝐿 + 𝛼 (4)

Where WML stands for “winners minus losers” and essentially states that by buying stocks that
recently have exhibited high returns (winners) and selling (shorting) stocks that recently have
exhibited poor returns (losers), investors can earn returns in excess of the market as a whole.
Carhart described the Momentum factor as the tendency for prices to continue rising if they are
going up and vice versa. He concluded that the Momentum factor successfully captures
significant excess returns from stocks that have recently performed well.

12



Piotroski (2000), Novy-Marx (2014), Fama and French (2014) and Asness et. al (2014) are
among some of the academic researchers that have studied the Quality factor. This factor
attempts to identify securities that exhibit certain characteristics that constitute ”high-quality”,
such as low leverage, high profitability and/or stable earnings (Piotroski, 2000). Asness et. al.
concluded that high-quality stocks delivered consistent superior risk-adjusted returns compared
to the market using a “quality-minus-junk” (QMJ) strategy, i.e. buying high-quality stocks and
selling low quality stocks. They defined quality based on different measurements of leverage,
growth and profitability.
Franzzini and Pedersen (2014) provided academic support for the outperformance of
stocks exhibiting low volatility in their paper titled “Betting Against Beta”. They found that low-
risk stocks outperformed high-risk stocks on a risk-adjusted basis over the long term. The
motivation behind this was that since investors like high returns, but are often not able to use
leverage, they instead overweight risky securities, which in turn pushes up their price thus
lowering their expected return.

3.2 Smart Beta

Even though research has been conducted on various factors that could explain excess returns
compared to traditional cap-weighted indices, the number of studies on Smart Beta is relatively
scarce and a definitive definition of Smart Beta is still vague. There are, however, recent
contributions that have helped explain the concept, namely the EDHEC (Ecole des Hautes
Etudes Commerciales du Nord)-Risk Institute in their paper “Alternative Equity Beta Investing:
A Survey” by Amenc, et al. (2015). This paper attempts to highlight potential benefits and risks
of Smart Beta strategies, as well as expanding on the rationale and empirical evidence for certain
Smart Beta factors including Value, Momentum, Low Risk, Size, Profitability and Investment.
Amenc and Goltz (2013) concluded that Smart Beta indices are likely to outperform cap-
weighted indexes but are exposed to several types of risks including factor tilts (systematic risk),
risks associated with specific inputs of a strategy and the risk of potentially severe
underperformance over long periods of time.
There is another element aside from academia that is worth mentioning when it comes to
Smart Beta. Research Affiliates LLC are one of the early adopters of Smart Beta asset allocation.
The company provides asset managers and large fund managers with strategy consulting
dedicated solely to various forms of Smart Beta. Hsu (2004) published on behalf of research
affiliates the paper “Cap-Weighted Portfolios are Sub-Optimal Portfolios”, where he argues that
capital pricing models may be misspecified. Research Affiliates’ presence within the field of Smart

13



Beta has influenced several global asset managers to provide investors the opportunity to invest
in Exchange Traded Funds (ETF’s) with Smart Beta exposure.

14



4. Factors
This section is dedicated to explaining and defining the Smart Beta factors analyzed in this thesis.
These factors stem from previous academic research within factor investing.

4.1 Market
Factors can be defined as a set characteristics or fundamentals of securities that are important in
explaining their performance and risk. According to the Capital Asset Pricing Model (CAPM)
developed in the 1960’s, stock performance is determined by one single factor, which is its
exposure to the market portfolio, otherwise known as Beta. In this thesis, we may also refer to
the market factor as the cap-weighted portfolio and will be used as a benchmark in terms of
performance. The two market portfolios used in this thesis are the S&P 500 and the Euro Stoxx
600 index (also referred to as Stoxx 600 throughout this paper).

4.2 Value
Since its first appearance as a rules-based strategy in the 1930’s (Graham and Dodd, 1934), the
Value factor has been used in various forms. The strategy aims at identifying undervalued
securities based on their price-to-fundamental ratios. Different types of measurements that have
been examined extensively in academia include ratios such as sales-to-earnings, price-to-cash
flows, price-to-earnings, price-to-book or book-to-market equity, as formalized by Fama and
French (1993).
Sanjoy and Basu (1977) constructed a portfolio consisting of securities that expressed low
price-to-earnings (P/E) ratios and found that they outperformed comparable indices over time.
The economic intuition of the Value factor requires the investor to assume that a relatively low
(high) fundamental value indicates that the asset is undervalued (overvalued), which is why
various measures of price-to-fundamentals are commonly used when capturing value. The
investor will typically invest in a portfolio of undervalued assets in relation to the market,
expecting the portfolio to outperform the corresponding index until levels are in line with the
rest of the market. The expected effect is therefore exponentially decreasing as it returns to
market standards. This suggests that most excess returns occur early in the business cycle. Sanjoy
and Basu also found that the Value factor exhibits a high level of sensitivity to the business cycle
over the long run.

The explanatory power of dividend as a Value factor can be linked to several theories. Firstly, the
Dogs of the Dow theory argues that dividends reflect the worth of a company and are relatively

15



stable over time, while stock prices vary over the business cycle. Proponents of this theory
therefore claim that high dividends relative to stock prices indicate the bottom of a business
cycle. Thus, investors can ride the wave of the business cycle by picking out securities with high
dividend-to-stock price ratios (Da Silva, 2001). The second theory follows a more behavioral line,
where high dividend stocks appear overvalued compared to stocks not paying dividend,
explained in the following example:

“Assume two, ceteris paribus, identical firms with no underlying growth:

Firm A follows a dividend policy of 100% of profits, while firm B decides not to pay any
dividend. In the beginning, both firms enjoy the same amount of profits, 100, and have a
share price of 1000, P/E is at 10 for both firms. One year later, profits remain at 100,
letting firm A report a P/E of 10, but the P/E for firm B is lower. Why? Firm B is
earning interest on the earnings from the first year, assuming an after-tax interest of 5%
nets firm B 105 in profit, causing P/E to fall to 9.5” (Clemens, 2012).

The scenario creates an optical illusion, which makes a dividend paying firm appear relatively
more expensive. The third theory attributes the explanatory power of dividend to differences in
taxation between wage and capital gains, where dividends in most economies provide an
alternative income less punished by taxes (Brennan, 1970).

4.2.1 Definition of the Value Smart Beta Factor

In this thesis, our Value Smart Beta factor will consist of four variables, including the price-to-
earnings ratio (P/E), price-to-book ratio (P/B), price-to-cash flow ratio (P/CF) and dividend
yield (these variables are further defined in Appendix B). Each security’s Value score will be
defined as the weighted average of their standardized variables (see Appendix A: Standardized
score). The value score for security i is calculated as follows:

𝑉𝑎𝑙𝑢𝑒 𝑆𝑐𝑜𝑟𝑒2 = −1 ∗ 0,25 ∗ 𝑍NO 2 + −1 ∗ 0,25 ∗ 𝑍NP 2 + −1 ∗ 0,25 ∗ 𝑍NQR 2 + 0,25 ∗ 𝑍S2;T 2 (5)

Where 𝑍NO 2 is the standardized score of security i’s P/E ratio, 𝑍NP 2 the standardized score of
security i’s P/B ratio, 𝑍NQR 2 the standardized score of security i’s P/CF ratio and 𝑍S2;T 2 the
standardized score of security i’s dividend yield. We multiply the first three standardized variables
by -1 in order to obtain a higher (lower) score for a lower (higher) fundamental to price ratio. The
top 100 value scores are selected and weighted according to our weighting scheme (section 5.2.2).

16



4.3 Size
Rolf Banz (1981) pioneered the Size factor, which targets comparatively smaller firms in order to
capture excess returns relative to larger firms. There are several theories as to why smaller
companies attain excess returns such as the exposure to default risk (Vassalou & Xing, 2004),
liquidity issues (Yakov, 2002) and financial distress (Chan & Chen, 1991) but none seem to fully
explain the increased returns earned by utilizing this investment form.
The main justification for size investing is the possibility of higher returns. Despite this,
small cap stocks are usually accompanied by a greater level of risk. According to Fama and
French (1993), small cap stocks have historically generated higher returns but the excess returns
are not free, as small cap stocks tend to exhibit a higher level of risk. Investors should therefore
be willing to ride out the “bad” times when size investing.
Exposure to the size factor is easily attained by using an equally weighted strategy, which
involves equally weighting all constituents on an index. The strategy provides relatively larger
exposure to the smaller firms, and less relative exposure to the larger firms. One thing to
consider when size investing is that smaller companies naturally have a higher default rate than its
larger counterparts making them more vulnerable to survivorship bias. Survivorship bias occurs
when analyzing past performance and selecting firms that have not defaulted (Rohleder et. al.,
2011). This essentially means neglecting firms that have defaulted during our sample, which in
turn could inflate the level of performance.

4.3.1 Definition of the size smart beta factor

In our thesis, all constituents on our benchmark indices will be weighted as follows:

𝑤2 = 1/𝑁 (6)

Where wi is the portfolio weight for security i, and N the total number of constituents on the
benchmark index.

4.4 Sharpe-Momentum
Carhart pioneered the Momentum factor with his four-factor model in the 1997 paper titled “On
Persistence in Mutual Fund Performance”. The factor is meant to play on trends, where stocks
with high recent returns tend to exhibit the same trend in the coming period and vice versa.
Extensive research had been conducted earlier on the Momentum factor with significant results
in terms of excess returns including Jegadeesh and Titman (1993) who monitored this factor on
17



the U.S. stock market between 1965 and 1989. They showed that the strategy of buying high
performing stocks and shorting losing stocks produced excess returns compared to the
benchmark index over the same period. Despite the historical success of the Momentum factor,
playing on trends is a short-horizon game. The effect reportedly dissipates in less than two years,
which is why Jegadeesh and Titman (1993) suggested three-to twelve-month holding periods for
securities. Momentum strategies therefore require frequent rebalancing, which could offset some
of the returns in trading costs.
Economists have been unable to settle on a single underlying theory to interpret the
success behind the momentum factor. Most research points towards behavioral economics,
where Barberis et. al. (1998) traced the Momentum effect to irrational decision making during
shorter periods due to news being incorporated slowly into prices. Overreactions in long series of
good (bad) news can cause the stock price to reach higher (lower) levels than otherwise justified.
The original Momentum strategy only considers historical returns, which could lead to
unwanted levels of risk. The Momentum strategy in this thesis therefore incorporates the Sharpe
ratio, designed by William Sharpe (1966). The Sharpe ratio was introduced as a measure of
reward-to-variability, which essentially captures the return in excess of the risk-free rate per unit
of risk. The Sharpe-Momentum factor in this thesis therefore builds on a security’s most recent
return performance, adjusted for the volatility during the same time period.

4.4.1 Definition of the Sharpe-Momentum Smart Beta factor

Instead of only using past 12-month return performance like Carhart, we will also incorporate
performance over the past six months and control for risk. We therefore define our Momentum
Smart Beta factor as the weighted average of the standardized six- and twelve-month trailing
Sharpe ratios. This is shown mathematically as follows:

𝑆ℎ𝑎𝑟𝑝𝑒 − 𝑀𝑜𝑚𝑒𝑛𝑡𝑢𝑚 𝑆𝑐𝑜𝑟𝑒2 = 0,5 ∗ 𝑍]^ _`"2a20b cd 2 + 0,5 ∗ 𝑍,.^ _`"2a20b cd 2 (7)

Where 𝑍]^ _`"2a20b cd 2 is the six-month trailing Sharpe ratio for security i and 𝑍,.^ _`"2a20b cd 2 is
the twelve-month trailing Sharpe ratio for security i. The top 100 Sharpe-Momentum scores are
selected and weighted according to our weighting scheme.

4.5 Low Volatility

According to the risk-return principle, security’s exhibiting low volatility should be associated
with lower returns. Despite this, countless academic papers confirm that a portfolio of stocks

18



with low volatility can outperform stocks with high volatility on a risk-adjusted basis and
therefore disjoint the SML as conveyed by the CAPM (Chan et. al., 1999; Haugen and Baker,
1991). The outperformance is one of the most persistent anomalies found in the field of
contemporary asset pricing but the lottery demand effect provides an explanation: High volatility
stocks are more likely to give a high, lottery-like payoff. Should riskier payoffs be preferred by a
majority of investors, the demand for high volatility stocks will be reflected in the price. The
lottery demand effect runs the hypothesis that investors exhibit risk-loving preferences when it
comes to the selection of stocks, and therefore low volatility stocks tend to be undervalued,
which increases their potential return (Bali et. al., 2016).
The acknowledgement of the low-volatility anomaly should naturally remove any excess
return it provides due to an increased number of investors following the strategy, but it has been
an established pattern for 90 years, and continues to outperform high volatility stocks in more
recent studies (Asness et. al., 2014).

4.5.1 Definition of Low Volatility Smart Beta factor

In this thesis, we define the Low Volatility Smart Beta factor as the one-year annualized volatility
(Further defined in section 5.3.2). The 100 stocks exhibiting the lowest one-year daily return
annualized volatility will be selected into the portfolio and weighted according to our weighting
scheme.

4.6 Quality
While research is largely indecisive on why quality companies tend to yield excess returns,
Campbell et. al. (2010) argued that cash flow fundamentals steer stock prices more than
macroeconomic variables, meaning a well-run firm can gain a competitive advantage through
careful capital management. This would in turn minimizes the risk of over-capitalization or over-
leveraging, which subsequently affects the stock price positively. Quality stocks tend to perform
better during bad times because if macroeconomic conditions start to deteriorate, more investors
will become risk-averse and start investing in stocks with sound capital management. This would
in turn push up the value of high quality stocks. This effect is the so called “flight-to-quality”
(Asness et al., 2014).

Joseph Piotroski (2000) approached the quality factor by selecting a portfolio of securities using a
broad accounting based fundamental analysis. Piotroski generated an F-Score, which determined
the financial strength of a company by the sum of nine binary variables:

19



𝐹cfg`h = 𝐹dij + 𝐹∆dij + 𝐹QRi + 𝐹jff`l"a + 𝐹∆^"`b20 + 𝐹∆_l`0 + 𝐹∆mh;h` + 𝐹∆m2nl2o + 𝐹Opqi%%h` (8)

Where 𝐹dij is given score 1 if return on assets is positive in the current year and 0 if
negative, 𝐹∆dij is given score 1 if the change in return on assets is higher in the current year than
the previous year and 0 otherwise, 𝐹QRi is given score 1 if cash from operations is positive in the
current year and 0 if negative, 𝐹jff`l"a is given score 1 if accruals are positive in the current year
and 0 if negative, 𝐹∆^"`b20 is given score 1 if the change in the growth margin is higher in the
current year than the previous year and 0 otherwise, 𝐹∆_l`0 is given score 1 if the asset turnover
ratio is higher in the current year in relation to the previous year and 0 otherwise, 𝐹∆mh;h` is
given score 1 if the leverage ratio is lower in the current year compared to the previous year and 0
otherwise, 𝐹∆m2nl2o is given score 1 if the Current ratio is higher in the current year and 0 it it’s
lower and 𝐹Opqi%%h` is given score 1 if new shares aren’t issued during the previous year and 0
otherwise. The highest quality stocks receive an F-Score of 9. This approach captures the
multidimensionality of the quality factor but requires a large amount of data.
There are more simple approaches to define quality such as Novy-Marx (2013) who
found that firms with high gross profitability earned returns in excess to the market benchmark
over longer periods. This factor was also employed by Fama and French (2014), who integrated it
into their 5-factor model. They defined quality as gross profitability divided total assets.

4.6.1 Definition of the Quality Smart Beta factor

Factors such as Momentum, Low Volatility and Value are relatively straightforward to
understand whereas the multidimensionality of the Quality factor makes it harder to clearly
define. Utilizing a simple, one-dimensional quality measurement such as Novy-Marx (2013) or
Fama and French (2014) would be sufficient from an academic perspective but adopting
Piotroski’s comprehensive approach would limit the selection of usable securities in our analysis.
We have therefore chosen to construct a Quality Smart Beta factor inspired by Piotroskis F-score
and the quality approach of Bender et. al (2014) with a few variable modifications. Instead of
assigning binary variables to our various quality variables like Piotroski, we use the standardized
score approach as in previous portfolios (further described in section 5.2). Bender et. al (2014)
argued that firms exhibiting low variability in earning-per-share over time can be perceived as
“high quality”. We therefore decided to incorporate this variable into our Quality Smart Beta
factor.
More formally, we define the Quality Smart Beta factor in this thesis as the weighted
average of the standardized scores of five variables. These include the debt-to-equity ratio to

20



capture the leverage factor; return on assets (ROA), return on equity (ROE) and operating cash
flow (CFO) capture the profitability factor complemented by earnings variability to capture
earnings quality. These variables are further defined in Appendix B. Our quality score for a
security, i, is calculated as follows:

𝑄𝑢𝑎𝑙𝑖𝑡𝑦 𝑆𝑐𝑜𝑟𝑒2 = (−0,2) × 𝑍SO d"u2g 2 + 0,2 × 𝑍QRi 2 + 0,2 × 𝑍dij 2 + 0,2 × 𝑍diO 2 + (−0,2) × 𝑍ONcv"` 2 (9)

Where 𝑍SO d"u2g 2 is the standardized score of security i’s debt-to-equity ratio, 𝑍QRi 2 the
standardized score of security i’s operational cash flow, 𝑍dij 2 the standardized score of security
i’s return on assets, 𝑍diO 2 the standardized score of security i’s return on equity and 𝑍ONcv"` 2 the
standardized score of security i’s earnings-per-share variability the past five years. The top 100
quality scores are selected into the portfolio and weighted according to our weighting scheme.

21



5. Thesis data and methodology

This section provides a description of the data and methodology used in this thesis. It starts by
describing the data followed by the portfolio generation section, which is intended to explain the
security selection, scoring and weighting process. This is then followed up by stating the various
performance measurements used in the empirical analysis

5.1 Thesis data

All data utilized in this thesis is collected from Bloomberg LP. Price data for all securities consists
of daily closing prices quoted in U.S. Dollars from 1994-01-01 to 2016-12-31 for the S&P 500
and in Euros for the Stoxx 600 index from 2002-01-01 to 2016-12-31, adjusted for dividends.
The S&P 500 is constructed to represent the broad domestic American economy through
changes in the aggregate market value of the 500 stocks, which represent all major industries. The
Euro Stoxx represents 600 large, mid and small cap companies across 18 different countries in
the European region (Bloomberg LP, 2017). Factor-specific data is collected using Bloomberg’s
Watchlist Analytics (WATC) tool, which provides information on key ratios and measurements
utilized in our security selection process. In rare cases, data on specific firms is interrupted by
corporate actions such as mergers, acquisitions or default. If a security lacks more than 80% of its
price data for a time period, it is excluded from the sample. Otherwise, the missing data is
replaced by an index average, corresponding to the return of the market portfolio.

5.2 Portfolio generation

5.2.1 Scoring
Each Smart Beta factor has a different number of variables attached to it. Our Quality factor, for
example, consists of data from five variables, including earnings variability, return on equity,
return on assets, cash flow from operations and the debt-to-equity ratio. All of these variables are
measured in different units, which creates a problem when attempting to generate a single factor-
specific score for each security. To forego this problem, we standardize each security’s respective
variable in order to normalize all variables. We then take the average of all standardized scores
for each variable to create a factor-specific score for each security (see Appendix A: Standardized
scores). Each security is then sorted by their factor specific score and the top 100 scores are
selected into the portfolio for weighting.

22



Graphic 1: Raw variable data is collected from Bloomberg for each security. The factors analyzed in this thesis are Value, Momentum,
Low Volatility, Size and Quality. We do not score the Size factor since it’s an equally-weighted strategy.

Graphic 2: Each security's variable is given a standardized score in order to normalize each variable.

The last step in the scoring process is to create a factor-specific score for each security. The
factor-specific score for each security is the average of its standardized scores:

cu"0o"`o2whovxy zcu"0o"`o2whovx{ z⋯z cu"0o"`o2whovx|

𝐹𝑎𝑐𝑡𝑜𝑟 𝑆𝑐𝑜𝑟𝑒2 =
}
(10)

23



Where 𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑𝑖𝑧𝑒𝑑 𝑉2, is the standardized score of variable 1 for security i, 𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑𝑖𝑧𝑒𝑑𝑉2.
the standardized score of variable 2 for security i and 𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑𝑖𝑧𝑒𝑑𝑉2w the standardized score of
variable Z for security i. This is then divided by Z, i.e., the number of variables associated with
the factor.

5.2.2 Weighting
An outlier is an observation that has been given an abnormally large score and therefore
threatens to skew the weights in the portfolio. Assigning exceptionally large weights to a single
security in a portfolio is not necessarily an issue, since it reflects a firm with exceptionally large
scores and should thus perform exceptionally well. The issue rather lies in that too much weight
threatens to consume any diversification gains in the portfolio, adding unnecessary risk
(Markowitz, 1952). While outliers are rare in our sample, we address the few cases by applying a
two-step framework for weighting the 100 securities in each portfolio. To limit the effect of
outliers we start by utilizing the Winsor method to make sure no factor-specific scores exceed a
certain threshold limit (Dixon, 1960) (see Appendix A: The Winsor method). In this thesis, we
Winsorize all factor-specific scores at ± 3 standard deviations from the mean. This means that if
scores are above (below) the Winsor threshold, those observations will be given a score which
represents three standard deviations above (below) the mean.

Figure 2: A normal distribution bell curve illustrating three standard deviations

from the mean. In our first weighting step, if observations exceed the Winsor
threshold, these observations are capped at three standard deviations.

To ease the weighting process, we assume the 100 securities in each portfolio is a large enough
data sample to approximately assume they follow a normal distribution (Kish, 1965). Using

24



normal distribution as an assumption in our sample is therefore justified as the potential bias it
induces is negligible.
The second step consists of complying with rules established by the European
Commission to ensure portfolios meet investor protection requirements in portfolio
diversification and liquidity. The Undertakings for the Collective Investment of Transferable
Securities (UCITS) limits individual weights in portfolios to five percent, and any exceeding
weights are redistributed evenly according to the standardized score of the individual firms (see
Appendix A: The UCITS framework).
The weight for each security in our portfolio is based on their factor-specific score and its
contribution to the sum of all 100 scores:

𝑍𝑤𝑖𝑛𝑠𝑜𝑟
𝑖
𝑤2 = 𝑛 𝑍𝑤𝑖𝑛𝑠𝑜𝑟
(11)
𝑖=1 𝑖

Where 𝑤2 denotes the weight for security i, and 𝑍2 denotes the Winsorized factor specific score
for security i.

5.3 Performance analysis

The performance of each Smart Beta strategy is measured against a benchmark index. Since the
underlying values of individual firms change over time, the portfolio must be rebalanced
frequently in order to maintain a portfolio that truly captures the desired factor. The portfolio
generation, including weighting, is repeated quarterly, i.e. every four months. The performance
measurements used are return, volatility, Sharpe ratio and excess return.

5.3.1 Portfolio Return

The cumulative return gathered as the return of asset i, 𝑅2 , times the weight attributed to that
asset, 𝑤2 , summed for all securities in the portfolio.

𝑅2N = ∑02‚, [𝑅2 ∗ 𝑤2 + 𝑅2z, ∗ 𝑤2z, + ⋯ + 𝑅2z0q, ∗ 𝑤2z0q, ] (12)

Where daily returns for the portfolio are measured as follows:

N… qN…†y
𝑅2 = (13)
N…†y

25



Where Pt is the closing price at time t and Pt-1 the closing price at time t-1.

5.3.2 Daily Return Annualized Volatility

In this thesis, we refer to the portfolio’s Daily Return Annualized volatility (or just Volatility) as
the degree of daily return variation over the course of a year. This is defined mathematically as:

N ˆ
𝜎j00l"a = s2 𝑥 252 (14)
‰ (` q`){
sˆ2 = xŠy …
‹q,
(15)

ˆ
Where s2 is the standard deviation of all daily returns observations during a year. We then
multiply this number by the square root of 252 to arrive at the daily return annualized volatility.

5.3.3 Sharpe ratio (SR)

The Sharpe Ratio is defined as the return in excess of the risk-free rate per unit of risk The
Sharpe ratio is mathematically defined as follows:

Œ
(dx qd• )
𝑆𝑅 = Œ (16)
s…

ˆ
Where 𝑅2 is the portfolio return, 𝑅% the risk-free rate of return as measured by the 3-month U.S.
ˆ
T-bill (Appendix E, Figure 27) and su the volatility of the portfolio.

5.3.4 Excess returns

In this thesis, we define excess returns as the difference in return between the portfolio and its
benchmark index. A portfolio’s return in excess of the market portfolio 𝑅hNŽ is calculated as:

𝑅hNŽ = 𝑅_N − 𝑅_^ (17)

Where 𝑅_N is the return of portfolio P during time period T and 𝑅_^ the return of the benchmark
index M during time T.

26



6. Empirical analysis
The following results are measured from January 1st, 1994 to December 31st, 2016 for the S&P
500 and January 1st, 2002 to December 31st, 2016 for the Euro Stoxx 600. We have examined the
performance of 5 different portfolios (Value, Momentum, Low Volatility, Size and Quality)
versus their respective benchmark index in terms of total return, risk-adjusted return, volatility
and excess returns.

6.1 Summary results

Smart Beta portfolios vs. S&P 500 - Cumulative Returns

300%

250%

200%

150%

100%

50%

0%

S&P 500 Value Momentum Low Volatility Size Quality

Figure 3: Cumulative returns for all portfolios, 1994 - 2016. S&P 500 in black for comparison.

Smart Beta portfolios vs. Stoxx 600 - Cumulative returns

150%

100%

50%

0%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

-50%

-100%

Stoxx 600 Value Momentum Low Volatility Size Quality

Figure 4: Cumulative returns for all portfolios, 2002 - 2016. Stoxx 600 in black for comparison.

27



Summary Statistics – Smart Beta portfolios vs. S&P 500

Measurement S&P 500 Value Momentum Low Vol Size Quality

Cumulative
196,56% 221,10% 286,03% 184,49% 239,29% 275,57%
return

Return* 8,55% 9,61% 12,44% 8,02% 10,40% 11,98%

Sharpe ratio* 0,59 0,58 0,65 0,62 0,65 0,75

Volatility* 17,07% 17,16% 19,14% 12,68% 17,53% 17,06%

Excess
- 1,07% 3,89% -0,53% 1,86% 3,44%
returns*
Table 1: Summary statistics for the smart beta portfolios on the S&P 500 from 1994 to 2016. S&P 500 included for comparison.
Worst (red) and best (green) values for each performance metric are highlighted. * Denotes average annual.

Summary Statistics – Smart Beta portfolios vs. Stoxx 600

Measurement Stoxx 600 Value Momentum Low Vol Size Quality

Cumulative
49,08% 82,56% 141,32% 102,51% 98,16% 102,77%
return
Return* 3,27% 5,50% 9,42% 6,83% 6,54% 6,85%

Sharpe ratio* 0,38 0,61 0,78 1,05 0,83 0,70

Volatility* 18,55% 17,95% 15,42% 10,14% 12,11% 14,97%

Excess
- 2,23% 6,15% 3,56% 3,27% 3,58%
returns*
Table 2: Summary statistics for the smart beta portfolios on the Stoxx 600 from 2002 to 2016. Stoxx 600 included for comparison.
Worst (red) and best (green) values for each performance metric are highlighted. * Denotes average annual.

From a total return perspective, it is clear that the best performing Smart Beta strategy over both
geographical regions was the Sharpe-momentum portfolio (Table 1, 2; Figure 3; 4). Both
portfolios managed to outperform the U.S and European benchmark indices by roughly 90 and
92 percentage points respectively. Despite this, the Sharpe-momentum portfolio had a hard time
outperforming the U.S benchmark index since 2006 as shown in the annual and cumulative
excess returns (Figure 9, 10). For the U.S portfolio, most of the outperformance came between
1999 and 2005 where it outperformed by roughly 108 percentage points. Since 2006, the
momentum portfolio has underperformed S&P 500 by about 15 percentage points and is also the
second worst performer post financial crisis (Figure 5).

28



Smart Beta portfolios vs. S&P 500 - Cumulative returns post financial crisis

160%

140%

120%

100%

80%

60%

40%

20%

0%
2009 2010 2011 2012 2013 2014 2015 2016

S&P 500 Value Momentum Low Volatility Size Quality

Figure 5: Cumulative returns post crisis for all portfolios, U.S.

Against the European benchmark, the outperformance in the Momentum portfolio has been
more consistent. It manages to not only be the best performer over the entire period but also
during the post financial crisis period, where it outperformed the benchmark by roughly 15
percentage points (Figure 6). As seen in figure 5 above, the size portfolio performs the best
against the S&P 500 post crisis, exhibiting a total excess return of roughly 32%.

Smart Beta portfolios vs. Stoxx 600 - Cumulative returns post financial crisis

100%

80%

60%

40%

20%

0%
2009 2010 2011 2012 2013 2014 2015 2016

Stoxx 600 Value Momentum Low Volatility Size Quality

Figure 6: Cumulative returns post crisis for all portfolios, Europe.

29



The only portfolio that did not outperform the U.S benchmark in terms of total return during the
whole period was the Low Volatility portfolio, which underperformed by roughly 12 percentage
points. Even though the portfolio underperforms in terms of return, it exhibits significantly
lower annualized volatility (12,68%) compared to the benchmark index (17,07%) and provides a
slight upgrade in risk-adjusted return (0,62 versus 0,59). Another positive aspect of the low
volatility portfolio is the evident trait of low drawdowns (Figure 3, 4). During bad economic
periods, the Low Volatility portfolios seem to outperform the index as seen during years 2000-
2002 and 2008 (Figure 11, 12). For the entire period, all of the portfolios managed to outperform
the European benchmark, both in terms of total and risk-adjusted return. The post crisis period
has seen only the Low Volatility portfolio underperform the benchmark index. The portfolio
underperformed by roughly 13%.
The best risk adjusted performer in the U.S was the Quality portfolio, which exhibited an
average annual Sharpe ratio of 0,75 compared to the S&P’s 0,59. It was also the second best
performer from a total return perspective (Table 1). The value portfolio exhibited the lowest
Sharpe ratio (0,58). This can most likely be attributed to its large drawdowns during poor macro
conditions since it outperformed the S&P 500 by about 23 percentage points over the whole
period.
The best risk-adjusted performer in Europe was the Low Volatility portfolio, which
exhibited an average annual Sharpe Ratio of 1,05 (Table 2). In terms of total return, it
outperformed the Euro Stoxx 600 index by about 53 percentage points over the entire period
and exhibits significantly less annualized volatility (10,14% versus 18,55%).
Figures 3 and 4 show clear signs of cyclical market trends for all portfolios, with peaks in early
2008 followed by large drawdowns due to the financial crisis. All smart beta portfolios showed
varying behavior between years 1999 and 2003. The Value and Low Volatility portfolios both
declined during the same time period between 1999 and 2000 whereas momentum and quality
seemed to follow the S&P 500’s trend, albeit with smaller drawdowns (Figure 3). The European
portfolios seem to exhibit peaks and troughs during the same time periods, with drawdowns in
late 2002, 2008 and 2015 (Figure 4).

6.2 Individual Portfolio results

6.2.1 Value
In terms of risk-adjusted returns, the value portfolio manages to outperform the benchmark
index in Europe but not in the U.S (Table 1, 2). The varying results in the two benchmark indices

30



could be explained by the underlying theory, which states that the value factor exhibits high
sensitivity to the overall macro environment (Sanjoy and Basu, 1977). Seeing that our data on the
European index starts in 2002 and the American equivalent begins in 1994, it is possible that the
timing of the investment played a role in the risk-adjusted outperformance of the Value portfolio
versus the Stoxx 600 index. This can be seen in figure 7, where the Value portfolio significantly
underperforms the S&P 500 from 1994 to the trough in late year 2000.

Cumulative excess returns - Value

40%

20%

0%

-20%

-40%

-60%

-80%
Figure 7: Cumulative excess returns for the Value portfolio versus the S&P 500 between 1994 and 2016.

Given that the U.S. and Europe roughly follow the same macroeconomic trends (Figure 3, 4),
this period of underperformance could have affected the Value portfolio in Europe as well if data
allowed us to measure from as early as 1994. Figure 8 below shows cumulative excess return for
the Value portfolio in Europe. What is clear from this figure is the significant underperformance
during the most recent financial crisis, which strengthens the argument that Value stocks tend to
underperform during poor macroeconomic conditions. Figures 16 and 17 in Appendix D show
that the Value portfolios exhibited higher levels of volatility during the most recent financial
crisis.

31



Cumulative excess returns - Value

60%

50%

40%

30%

20%

10%

0%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Figure 8: Cumulative excess returns for the Value portfolio versus the Euro Stoxx 600 between 2002 and 2016.

The overall results for the Value portfolio are in line with Fama and French (1993) who found
that Value stocks tend to outperform the market as a whole over the long run (Table 1, 2). When
studying figures 7 and 8, it seems as if the Value portfolio performs well during the early phases
of the business cycle corresponding to the results of Sanjoy and Basu (1977). This is indicated by
the outperformance from 2002-2005 and 2009-2011 across both geographical regions. The Value
portfolio exhibited higher returns versus the S&P 500 in 61 of the 92 quarters examined, and in
37 of the 60 quarters on the European equivalent (Appendix C: Table 11, 19).

6.2.2 Sharpe-Momentum
The Momentum portfolios are the most impressive in the data set in terms of total return during
the entire period. In 67 out of the 92 quarters examined, the Momentum portfolio outperformed
the return of the S&P 500, and in 41 of 60 quarters on the European equivalent (Appendix C:
Tables 11; 19). Seeing that Behavioral Economics attributes the momentum anomaly to irrational
investor behavior i.e., overreaction to good or bad news, one could argue that our Momentum
portfolios should exhibit much larger underperformance versus the benchmark than depicted in
Figures 9 and 10. The Momentum portfolios also exhibited lower volatility during the most
recent financial crisis (Appendix D, Figures 19, 20). A particularly interesting result from
analyzing both Momentum portfolios are the differing results between years 2006 and 2016. The
Momentum portfolio in the U.S clearly stagnates in terms of excess returns during this period but
the outperformance remains relatively consistent in Europe.

32



Cumulative excess returns - Momentum

140%

120%

100%

80%

60%

40%

20%

0%

Figure 9: Cumulative excess returns for the Momentum Portfolio versus the S&P 500 between 1994 and 2016.

Cumulative excess returns - Momentum

120%

100%

80%

60%

40%

20%

0%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Figure 10: Cumulative excess returns for the Momentum portfolio versus the Euro Stoxx 600 index between 2002 and 2016.

Nevertheless, our results for the entire period are in line with Carhart (1997) as well as Jegadeesh
and Titman (1993), who found that momentum stocks tend to exhibit excess returns compared
to the market over the long run.

33



6.2.3 Low Volatility

Previous academic studies such Frazzini and Pedersen (2014), Chan et. al. (1999) and Haugen
and Baker (1991) suggest that low volatility stocks tend to outperform high volatility stocks on a
risk-adjusted basis because they are systematically undervalued due to risk-loving preferences
among investors. This is in line with our results as both Low Volatility portfolios exhibit
superior risk-adjusted returns relative to their benchmark indices. The portfolios also behaved
according to expectations when it comes to volatility, where both portfolios exhibited the lowest
annualized volatility over the entire period (Table 1, 2). Table 5 and 9 (Appendix C) shows the
Low Volatility portfolio in the U.S. exhibiting the lowest daily return annualized volatility in every
year between 1994 and 2016, with similar results for the European portfolio (See Appendix D for
the annualized volatility of both Low Volatility portfolios throughout the time period). Another
significant result is the outperformance during periods of economic turmoil such as the most
recent financial crisis as seen in Figures 11 and 12 below.

Cumulative excess returns - Low Volatility

80%

70%

60%

50%

40%

30%

20%

10%

0%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Figure 11: Cumulative excess returns for the Low Volatility portfolio versus the Stoxx 600 between 2002 and 2016.

34



Cumulative excess returns - Low Volatility

10%
0%
-10%
-20%
-30%
-40%
-50%
-60%
-70%
-80%
-90%
Figure 12: Cumulative excess returns for the Low Volatility portfolio versus the S&P 500 between 1994 and 2016.

In 62 out of the 92 quarters examined, the Low Volatility factor outperformed the return of the
S&P 500, and in 32 of 60 quarters on the European equivalent (Appendix C: Tables 11, 19).

6.2.4 Size
Both Size portfolios outperform their respective benchmark indices in terms of total and risk-
adjusted returns, which is in line with Fama and Fench (1993) who concluded that small cap
stocks tend to outperform large cap stocks over the long run. Fama and French also established
that even though smaller companies tend to outperform the market as a whole, they are exposed
to greater levels of risk. This is evident in the U.S. but not in Europe, as the Size portfolio
exhibited lower volatility during for the entire period (Table 1, 2). A reason for this could be the
diversification effect as suggested by Markowitz (1952), since risk is spread out equally overall
index constituents (See Appendix D for the daily return annualized volatility of both Size
portfolios throughout the time period). As depicted in figure 13 and 14 below, both Size
portfolios seem to exhibit underperformance in relation to their benchmark indices during poor
macroeconomic conditions i.e., during the dot-com bubble of the early 2000’s in the U.S. as well
as the most recent financial crisis.

35



Cumulative excess returns - Size

50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Figure 13: Cumulative excess returns for the Size portfolio versus the Stoxx 600 between 2002 and 2016.

Cumulative excess returns - Size

50%
40%
30%
20%
10%
0%
-10%
-20%
-30%
-40%
-50%
Figure 14: Cumulative excess returns for the Size portfolio versus the S&P 500 between 1994 and 2016.

Both Size portfolios perform particularly well in the post crisis period as seen in Figures 13 and
14 above. The outperformance during this period was about roughly 30% versus the S&P 500
and 12% versus the Stoxx 600 index. The Size portfolio outperformed the S&P 500 in 66 out of
the 92 quarters examined. The corresponding result for Europe was 35 of 60 quarters (Appendix
C: Tables 11, 19).

36



6.2.5 Quality
In line with the results of Asness et. al. (2014), the Quality portfolio exhibits superior risk-
adjusted returns versus both benchmark indices (Tables 1, 2). As mentioned in section 4, the
underlying reasons why quality stocks tend to outperform the market in risk-adjusted terms are
unclear but Campbell et. al. (2010) argued that cash flow fundamentals steer stock prices more
than macroeconomic variables meaning that a well-run firm can gain a competitive advantage
through careful capital management. This would in turn minimize the risk of over-capitalization
or over-leveraging, which subsequently affects the stock price positively.

Cumulative excess returns

70%

60%

50%

40%

30%

20%

10%

0%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Figure 15: Cumulative excess returns for the Quality portfolio versus the Stoxx 600 between 2002 and 2016.

According to Asness et al. (2014), quality stocks tend to perform better during bad times because
if macroeconomic conditions start to deteriorate, more investors will become risk-averse and
start investing in less risky stocks. This would in turn push up the value of high-quality stocks.
This effect is the so called “flight-to-quality” effect and is evident for both Quality portfolios in
our analysis. This can be seen in figures 14 and 15 where both portfolios outperform their
respective benchmark index during the most recent financial crisis as well as during the dotcom
bubble of the early 2000’s in the U.S. The results are also consistent with Asness et. al. (2014,
who concluded that Quality stocks tend to outperform the market on a risk-adjusted basis over
the long run.

37



Cumulative excess returns

100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%

Figure 16: Cumulative excess returns for the Quality portfolio versus the S&P 500 between 1994 and 2016.

38



6.2.6 Portfolio correlation

Correlation Matrix
S&P 500 Value Momentum Low Volatility Size Quality
S&P 500 1
Value 0,872231 1
Momentum 0,885352 0,728588586 1
Low Volatility 0,868248 0,869466357 0,765231016 1
Size 0,018790 0,033352938 0,026106398 0,000814169 1
Quality 0,973525 0,84195112 0,856633999 0,851596052 0,010752 1
Table 3: Correlation matrix for all portfolios, U.S.

Correlation Matrix
Stoxx 600 Value Momentum Low Volatility Size Quality
Stoxx 600 1
Value 0,67669232 1
Momentum 0,721489588 0,683531958 1
Low Volatility 0,768505838 0,755981667 0,831248752 1
Size 0,516431069 0,544309488 0,523920165 0,502933879 1
Quality 0,819895803 0,822904911 0,856664972 0,884802083 0,565236191 1
Table 4: Correlation matrix for all portfolios, Europe.

As seen in tables 3 and 4, the Quality Smart Beta portfolio has the highest correlation with both
benchmark indices. This could be explained by the underlying variables that we use to define our
Quality factor (section 4.6). These variables are common characteristics among large cap stocks
and considering that benchmark indices are market-cap weighted, the high correlation between
Quality and both benchmark indices should be justified. The Size portfolios tend to exhibit the
lowest correlation with the Quality and Low Volatility portfolios. This is in line with Fama and
French (1993) who established that even though the Size factor tends to outperform the market
as a whole, they tend to exhibit a higher level of risk. It is therefore plausible that the Size
portfolio exhibits the lowest correlation with the Low Volatility portfolios. The correlation
matrices show differing results from a geographical perspective. Besides the fact that we have a
vast difference in the number of observations in each data set, both benchmark indices are also
fundamentally different. The S&P 500 is constructed to represent the entire American economy
whereas the Stoxx 600 represents 18 different countries in Europe (section 5.1). We can
therefore expect to have a slight variance in portfolio behavior for each geographical region.

39



7. Conclusion
Based on the empirical analysis carried out in this thesis, we can conclude that 9 out of 10 Smart
Beta portfolios outperformed their respective benchmark index on a risk-adjusted basis. This
result suggests that exposure to Smart Beta factors can in fact lead to significant improvements in
risk-adjusted returns. When attempting to analyze our results through the lens of academic theory
and literature, we find that Smart Beta investing doesn’t seem to always abide by the rules. This is
particularly true when taking into account the widely accepted theory of efficient markets, which
essentially implies that cap-weighted indices are mean-variance efficient, i.e. offer the best return-
to-risk ratio given a specified level of risk tolerance. According to our results, exposure to Smart
Beta factors has historically proven to be beneficial, both in total and risk-adjusted returns, which
is in line with several previous academic studies within the subject.
The question if markets truly are efficient from an asset allocation perspective is highly
dependent on the sample period analyzed. This is evident in our data set as we see some
portfolios clearly stagnating in performance when examining the post crisis period. When
drawing conclusions based on historical data, timing is certainly one of the main factors. The
portfolios also exhibit different results across the two geographical regions we examine. As
mentioned earlier, The S&P 500 and the Euro Stoxx 600 indices are fundamentally different in
what they represent and thus, the differing results are to be expected. Another problem when
analyzing historical data is the survivorship bias. This may have skewed the data set, especially
when analyzing Smart Beta factors exposed to illiquid firms. In the real world, exposure to firms
that default could be associated with costs that we cannot control for properly in our analysis.
Since this paper compares Smart Beta strategies with passively investing in a cap-weighted index,
one must account for the cost of active management and brokerage fees. While the brokerage
fees of today do not offset the historical outperformance of Smart Beta strategies, the time and
costs associated with actively constructing and rebalancing portfolios could. This may provide
one of many explanations as to why markets remain inefficient given that investors face different
financial and informational constraints, creating an uneven playing field.
The clear historical outperformance of Smart Beta strategies definitely questions if
markets are efficient. Whether the explanation for this is behavioral or economic, our conclusion
is that it’s possible to outperform the market as a whole with exposure to certain risk factors over
the long run. Despite this, we advise investors to be aware of the costs associated with Smart
Beta investing and to remember that past performance is not an indicator of future results.

40



7.1 Suggested further research

The data set for the S&P 500 stretches back to 1994, while the corresponding data set for the
Euro Stoxx 600 only goes back to 2002. A larger data set for the European index would yield
more specified results, allowing for a more accurate comparison across geographical regions. We
therefore suggest our method of portfolio evaluation to be conducted when more data is
available.
In order to find out what drives each Smart Beta factor, further research should be
conducted on how each Smart Beta portfolio performs in various stages of the business cycle.
This can be done by regressing each portfolio on specific macroeconomic indicators such as
GDP growth, inflation or unemployment.

41



References
Altman, Douglas G., and J. Martin Bland. 1995. "Statistics notes: the normal distribution." Bmj 310.6975:
298.

Amenc, N., et al. 2015. "Alternative Equity Beta Investing: A Survey.". An EDHEC-Risk Institute
Publication.

Amenc, Noël, and Goltz, Felix. 2013. "Smart Beta 2.0.". The Journal of Index Investing 4.3: 15-23.

Amihud, Yakov. 2002. "Illiquidity and stock returns: cross-section and time-series effects.". Journal of
financial markets 5.1: 31-56.

Ang, Andrew. 2014. “Asset management: A systematic approach to factor investing.”. Oxford University
Press.

Asness, Clifford S., Andrea Frazzini, and Lasse Heje Pedersen. 2014 "Quality minus junk.".

Bali, Turan G., et al. 2016. "A lottery demand-based explanation of the beta anomaly.". Journal of Financial
and Quantitative Analysis.

Banz, Rolf W. 1981. "The relationship between return and market value of common securities." Journal of
financial economics 9.1:3-18.

Barberis, Nicholas, Andrei Shleifer, and Robert Vishny. 1998."A model of investor sentiment." Journal of
financial economics 49.3: 307-343.

Basu, Sanjoy. 1977. "Investment performance of common securities in relation to their price-earnings
ratios: A test of the efficient market hypothesis." The journal of Finance 32.3: 663-682.

Bender, Jennifer, Eric Brandhorst, and Taie Wang. 2014. "The Latest Wave in Advanced Beta: Combining
Value, Low Volatility, and Quality." The Journal of Index Investing 5.1: 67-76.

Black, Fischer. 1993. "Beta and return." The journal of portfolio management 20.1: 8-18.

Bowen Jr, John J., and David G. Booth. 1993. "Active versus Passive Investing." The Journal of Investing 2.1:
11-13

Brennan, Michael J. 1970. "Taxes, market valuation and corporate financial policy." National tax
journal 23.4: 417-427.

Campbell, John Y., Christopher Polk, and Tuomo Vuolteenaho. 2010 "Growth or glamour?
Fundamentals and systematic risk in stock returns." Review of Financial Studies 23.1: 305-344.

Carhart, Mark M. 1997. "On persistence in mutual fund performance." The Journal of finance 52.1: 57-82.

Chan, Louis KC, Jason Karceski, and Josef Lakonishok. 1999. "On portfolio optimization: Forecasting
covariances and choosing the risk model." Review of Financial Studies 12.5: 937-974.

Chan, K. C., and Nai-Fu Chen. 1991. "Structural and return characteristics of small and large firms." The
Journal of Finance 46.4: 1467-1484.

Clemens, Michael. 2012. "Dividend investing: Strategy for long-term outperformance." BankInvest Asset
Management A/S, Sundkrogsgade 7, DK-2100 Copenhagen, Denmark.

42



Da Silva, Andre LC. 2001. "Empirical tests of the Dogs of the Dow strategy in Latin American stock
markets." International Review of Financial Analysis 10.2: 187-199.

Dixon, W. J. 1960. “Simplified Estimation from Censored Normal Samples”, The Annals of Mathematical
Statistics, 31: 385–391.

Fama, Eugene F. 1965. "The behavior of stock-market prices." The journal of Business 38.1: 34-105.

Fama, Eugene F. 1991. "Efficient capital markets: II." The journal of finance 46.5: 1575-1617.

Fama, Eugene F., et al. 1969. "The adjustment of stock prices to new information." International economic
review 10.1: 1-21.

Fama, Eugene F., and Kenneth R. French. 1988. "Dividend yields and expected stock returns." Journal of
financial economics 22.1: 3-25.

Fama, Eugene F., and Kenneth R. French. 1993. "Common risk factors in the returns on securities and
bonds." Journal of financial economics 33.1: 3-56.

Fama, Eugene F., and Kenneth R. French. 2016. "Dissecting anomalies with a five-factor model." Review of
Financial Studies 29.1: 69-103.

Frazzini, Andrea, and Lasse Heje Pedersen. "Betting against beta. 2014. "Journal of Financial
Economics 111.1: 1-25.

Gordon, Myron J., and Eli Shapiro. 1956. "Capital equipment analysis: the required rate of
profit." Management science 3.1: 102-110.

Graham, Benjamin, and David L. Dodd. 1934. “Security analysis: Principles and technique.”. McGraw-Hill,
New York.

Grossman, Sanford J., and Joseph E. Stiglitz. 1980. "On the impossibility of informationally efficient
markets." The American economic review 70.3: 393-408.

Haugen, Robert A., and Nardin L. Baker. 1991. "The efficient market inefficiency of capitalization–
weighted stock portfolios." The Journal of Portfolio Management 17.3: 35-40.

Hsu, Jason C. 2004. "Cap-weighted portfolios are sub-optimal portfolios." Journal of Investment
Management 4.3: 1-10.

Jegadeesh, Narasimhan, and Sheridan Titman. 1993. "Returns to buying winners and selling losers:
Implications for stock market efficiency." The Journal of finance 48.1: 65-91.

Kish, Leslie. 1965. "Sampling organizations and groups of unequal sizes." American sociological review: 564-
572.

Lintner, John. 1965. "The valuation of risk assets and the selection of risky investments in stock portfolios
and capital budgets." The review of economics and statistics, 65:13-37.

Malkiel, Burton G. 2003. "The efficient market hypothesis and its critics." The Journal of Economic
Perspectives 17.1: 59-82.

Markowitz, Harry. 1952. "Portfolio selection." The journal of finance 7.1 52: 77-91.

Mossin, Jan. 1966. "Equilibrium in a capital asset market." Econometrica: Journal of the econometric society 66:
768-783.

43



Novy-Marx, Robert. 2013. "The other side of value: The gross profitability premium." Journal of Financial
Economics 108.1: 1-28.

Piotroski, Joseph D. 2000. "Value investing: The use of historical financial statement information to
separate winners from losers." Journal of Accounting Research 00: 1-41.

Rohleder, Martin, Hendrik Scholz, and Marco Wilkens. 2011. "Survivorship bias and mutual fund
performance: Relevance, significance, and methodical differences." Review of Finance 15.2: 441-474.

Ross, Stephen A. 1976. "The arbitrage theory of capital asset pricing." Journal of economic theory 13.3: 341-
360.

Schiller, Robert J. 2000. "Irrational exuberance." Princeton UP.

Sharpe, William F. 1964. "Capital asset prices: A theory of market equilibrium under conditions of
risk." The journal of finance 19.3: 425-442.

Sharpe, William F. 1966. "Mutual fund performance." The Journal of business 39.1: 119-138.

Sharpe, William F., Gordon J. Alexander, and Jeffery V. Bailey. 1999. “Investments”. Vol. 6. Upper Saddle
River, NJ: Prentice-Hall.

Treynor, Jack L. 1961. "Toward a theory of market value of risky assets." Unpublished manuscript 6.

Vassalou, Maria, and Yuhang Xing. 2004. "Default risk in equity returns." The Journal of Finance 59.2: 831-
868.

TIA, The Investment Association. Available [Online] 2017-05-20:

http://www.investinginfunds.org/facts-about-funds/how-uk-funds-are-regulated/investment-limits-
what-fund-managers-can-invest-in.html

NYSE, New York Stock Exchange. Available [Online]2017-05-20:

http://www.nyxdata.com/nysedata/asp/factbook/viewer_edition.asp?mode=table&key=2206&category
=4

RA, Research Affiliates LLC. Available [Online]2017-05-20:

https://www.researchaffiliates.com/en_us/about-us.html

44



Appendix A
Standardized score
We rely on data of large indices carrying over 500 different firms. The large data pool allows us to
assume that the values corresponding to individual firms follow a normal distribution* centered
around a mean, µ, forming the classic bell-shaped curve below. Note that normal distribution
implies that 68 percent of observations are within one standard deviation, 95 percent within two
standard deviations and 99.7 percent are within three standard deviations from the mean (Altman
and Bland, 1995).

Figure 3: The normal distribution with centered mean and standard deviation ticks.
See Altman and Bland (1995).

The standardized score is computed in two steps: First, the geometric mean of each underlying
variable is calculated from the total pool of observations. This gives us an estimated center-value
that will be used as the comparable value. Secondly, we assign individual standardized scores, 𝑧2 ,
to each underlying variable of interest by subtracting the underlying variable mean from the
individual value, 𝑥. The last step involves dividing the difference between the underlying variables
value and the mean value by the standard deviation of all underlying variables. This procedure
gives us a standardized number, allowing us to treat all underlying variables similarly, regardless
of any discrepancies in units.

•x qµ
𝑧2 = (18)
•

,
All standardized scores for each underlying variable are given equal weights, , and added
0

together to compile an overall standardized factor specific score, Z.

*
Normal distribution is also known as Gaussian distribution (Altman and Bland, 1995).
45



,
𝑍2 = (𝑧2 + 𝑧2z, + 𝑧2z. + ⋯ + 𝑧0 ) (19)
0

We now possess a large data set of firms with corresponding overall standardized scores based
on selected underlying variables predetermined to reflect a Smart Beta factor. Lastly, the data set
is sorted by scores, where the firms with the highest scores are on top of the list. The top 100
firms are selected to proceed to the weighting process, where it will be determined how much
weight each firm will have in the final portfolio. The firms with scores below the top-100 are
discarded.

The UCITS framework

An outlier is an observation that has been given an abnormally large score and therefore
threatens to skew the weight in the portfolio. Assigning exceptionally heavy weights to a single
firm in a portfolio is not necessarily an issue, since it reflects a firm with exceptionally large
scores, and should thus perform exceptionally well. The issue rather lies in that too much weight
threatens to consume any diversification gained in the portfolio, adding unnecessary risk. There
are a few alternatives to solve this issue, but as in any data-altering practice, it induces bias in the
sample. We therefore face a tradeoff between the vulnerability of outliers and bias. The path
chosen in this paper is to accept the bias and follow the regulatory framework UCITS,
established by the European Commission which is used by fund managers in the financial
industry to ensure the portfolio meets investor protection requirements in portfolio
diversification and liquidity (TIA, 2016).

The Undertakings for the Collective Investment of Transferable Securities (UCITS) states:
1. UCITS can invest in an absolute minimum of 16 assets: 4 holdings of up to 10% each
plus 12 holdings of up to 5% each.

2. A UCITS fund may invest no more than 5% of its value in approved securities or money
market instruments issued by any one body. This limit can be increased to 10% provided
that the total value of any holdings between 5% and 10% does not exceed 40% of the
fund.

3. No more than 20% of the fund as deposits with any one bank

46



4. No more than 20% of the fund invested in any one other fund

5. Up to 35% of the fund in any one bond issue provided the rest of the fund is invested in
other types of assets; or a minimum of six issues if the fund is over 35% invested in
Government bonds.

6. No more than 10% exposed in derivatives with another bank as counterpart

7. Hold no more than 20% of the voting shares of a company

8. Hold no more than 10% of the bonds issued by a company

9. Hold no more than 20% of the value of another fund

(TIA, 2016)

Since our data consist solely of securities, we can disregard bullet points 3-to-9, leaving us to
adjust the portfolio according to rules 1 and 2. Rule one regulates diversification by setting a
minimum number of securities in a portfolio, which we comply with by using portfolios of 100
securities. The UCITS rules thereby limits us through the second rule, which also regulates
diversification, but by controlling for amount invested in a single entity. In cases of extreme
outliers, our method will have to adhere to the UCITS 2nd rule, by allowing investments of no
more than 5 percent in one stock.

The Winsor method

Applying the UCITS limitation to the portfolio raises another dichotomy. Either exclude outliers
above the 5 percent threshold from the data set or invest exactly up until the five percent level
and redistribute the excess funds. The first method is commonly known in statistics as trimming
and is used when an outlier is considered a true anomaly which does not at all belong in the
distribution. Our outliers are firms that have scored exceptionally well, we therefore wish to keep
them without hurting the portfolio’s diversification properties.
We apply two-step framework, a first step framework using the Winsor method, which is
ultimately regulated by a second, framework consisting of the UCITS rules. Winsor is an applied
statistic method designed to limit the effect of outliers by adjusting their value to an outer limit
(Dixon, 1960). Our data set is Winsorized on the overall standardized scores of individual firms,

47



and the limit is set to target scores that exceed three standard deviations from the mean. Finally,
individual weights are limited to five percent, as by the UCITS framework, and any exceeding
weights are redistributed evenly according to the standardized score of the individual firms.
The weight, i.e. amount invested in each firm is based on their score and its contribution
to the sum of all scores:

𝑍𝑤𝑖𝑛𝑠𝑜𝑟
𝑖
𝑤2 = 𝑛 𝑍𝑤𝑖𝑛𝑠𝑜𝑟
(20)
𝑖=1 𝑖

Where 𝑤2 denotes the weight for security i, and 𝑍2 , denotes the factor specific score for security i.

48



Appendix B
Value variables
P/E Ratio
The price-to-earnings ratio (P/E) is defined as the ratio of a security’s share price to its earnings
per share (EPS). This is defined mathematically as:

N c‘"`h ˆ`2fh
= (21)
O O"`020b8 ˆh` 8‘"`h

P/B Ratio
The price-to-book ratio is defined as the ratio between a security’s current share price and it book
value. This can also be referred to as the Market-to-book ratio. This is defined mathematically as:

N c‘"`h ˆ`2fh
= (22)
P Pgg’ ;"alh

P/CF Ratio
The price-to-cash flow ratio is the ratio between a security’s share price and its operating cash
flow. This is defined mathematically as:

N c‘"`h ˆ`2fh
= (23)
QR iˆh`"u20b f"8‘ %ag“

Dividend yield
A security’s dividend yield or dividend price ratio is defined as ratio between a security’s
dividends per share and its share price. This is defined mathematically as:

S2;2oh0o8 ˆh` 8‘"`h

𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑 𝑦𝑖𝑒𝑙𝑑 = (24)
c‘"`h ˆ`2fh

Quality Variables
Return on assets (ROA)
Return on assets is the ratio between a firm’s net income and total assets. ROA is an indicator of
how profitable a firm is in relation to its assets. This is defined mathematically as:

‹hu 20fg)h
𝑅𝑂𝐴 = (25)
_gu"a "88hu8

49



Return on Equity (ROE)

Return on Equity is the ratio between a firm’s net income and shareholders’ equity. This
profitability measurement reveals how much a company generates in profit in relation to the
amount of money shareholders have invested. This is defined mathematically as:

‹hu 20fg)h
𝑅𝑂𝐸 = (26)
c‘"`h‘gaoh`8 ™ hnl2uš

Cash flow from operations (CFO)

Cash flow from operations or operating cash flow is a company’s cash generation from revenue
excluding costs such as investment in securities or long-term investment in capital. CFO is
defined mathematically as:

𝐶𝐹𝑂 = 𝐸𝐵𝐼𝑇 + 𝐷𝑒𝑝𝑟𝑒𝑐𝑖𝑎𝑡𝑖𝑜𝑛 − 𝑇𝑎𝑥𝑒𝑠 ± ∆𝑊𝑜𝑟𝑘𝑖𝑛𝑔 𝑐𝑎𝑝𝑖𝑡𝑎𝑙 (27)

Debt-to-equity ratio (D/E Ratio)

The D/E ratio is defined as the relationship between a company total debt and total equity,
which measures a firm’s financial leverage. This is shown mathematically as:

S _gu"a oh¡u
= (28)
O _gu"a hnl2uš

Earnings per share variability (EPS Variability)

Earnings per share variability is defined as the standard deviation of the change in earnings per
share growth during a period. In this thesis, we use EPS variability for the previous 5 years. This
is defined mathematically as:

𝜎∆ONc ¢`g“u‘ _q,:_q¤ (29)

50



Appendix C
Annual winners and losers
a) S&P 500

Annual returns - Smart Beta portfolios vs. S&P 500

Year S&P 500 Value Momentu Low Vol Size Quality
1994 -1,07% -8,00% m
-0,41% -5,28% -1,12% 8,21%
1995 29,65% 21,89% 30,05% 27,80% 25,12% 33,99%
1996 19,10% 11,23% 19,73% 12,55% 17,89% 24,00%
1997 28,71% 23,56% 26,77% 24,43% 20,98% 31,40%
1998 25,64% 10,50% 22,74% 6,05% 13,76% 32,21%
1999 19,42% -8,30% 43,09% -10,79% 8,42% 20,56%
2000 -8,19% 22,10% 25,69% 18,34% 8,67% 3,50%
2001 -11,67% 22,95% -9,74% 1,85% 1,66% -1,85%
2002 -23,20% -9,93% -17,57% -5,57% -15,35% -12,25%
2003 24,89% 24,09% 45,97% 18,73% 31,91% 23,16%
2004 9,26% 13,69% 17,96% 13,27% 15,80% 8,82%
2005 3,50% 2,79% 16,44% 0,61% 8,83% 3,72%
2006 13,25% 19,55% 9,19% 14,85% 12,39% 13,42%
2007 4,67% -6,19% 9,20% 0,03% 0,50% 12,91%
2008 -40,18% -35,08% -49,02% -23,80% -42,52% -36,45%
2009 24,66% 41,86% 24,76% 16,49% 38,49% 31,96%
2010 13,60% 16,26% 12,20% 9,74% 21,25% 13,41%
2011 2,71% 3,51% -0,07% 10,71% 1,85% 7,67%
2012 13,44% 9,76% 8,85% 7,30% 15,45% 10,40%
2013 26,53% 20,23% 32,16% 19,54% 28,60% 27,79%
2014 11,41% 10,78% 10,31% 15,33% 14,43% 12,88%
2015 0,46% -15,49% 4,50% 3,20% -2,47% -2,10%
2016 9,97% 29,35% 3,21% 9,12% 14,73% 8,20%
Table 3: Yearly annual returns for the smart beta portfolios on the S&P 500 from 1994 to 2016. S&P 500 included for comparison.
Worst (red) and best (green) values for each year are highlighted.

51



Sharpe ratios - Smart Beta portfolios vs. S&P 500

Year S&P 500 Value Momentum Low Vol Size Quality
1994 -0,55 -1,22 -0,40 -1,10 -0,57 0,38
1995 3,08 2,10 2,28 3,45 2,55 2,93
1996 1,19 0,63 1,15 0,77 1,20 1,41
1997 1,30 1,48 1,05 1,57 1,04 1,39
1998 1,03 0,42 0,83 0,09 0,48 1,38
1999 0,81 -0,92 1,72 -1,25 0,25 0,91
2000 -0,64 0,84 0,56 0,73 0,14 -0,12
2001 -0,70 1,09 -0,72 -0,12 -0,09 -0,23
2002 -0,96 -0,45 -0,89 -0,38 -0,64 -0,58
2003 1,40 1,46 2,33 1,45 1,71 1,38
2004 0,71 1,25 1,15 1,35 1,16 0,70
2005 0,03 -0,04 0,93 -0,27 0,50 0,05
2006 0,84 1,67 0,30 1,29 0,69 0,77
2007 0,01 -0,60 0,25 -0,31 -0,24 0,54
2008 -1,02 -0,74 -1,27 -0,85 -1,02 -0,92
2009 0,90 0,95 0,86 0,97 1,11 1,23
2010 0,75 0,98 0,53 0,76 1,04 0,83
2011 0,11 0,14 0,00 0,65 0,07 0,34
2012 1,05 0,87 0,67 0,87 1,04 0,75
2013 2,40 1,79 2,38 1,91 2,34 2,49
2014 1,00 1,06 0,74 1,66 1,30 1,11
2015 0,03 -0,95 0,29 0,23 -0,16 -0,14
2016 0,74 1,65 0,23 0,80 0,95 0,58
Table 4: Yearly Sharpe ratios for the smart beta portfolios on the S&P 500 from 1994 to 2016. S&P 500 included for comparison.
Worst (red) and best (green) values for each year are highlighted.

52



Annualized volatility - Smart Beta portfolios vs. S&P 500

Year S&P 500 Value Momentum Low Vol Size Quality
1994 9,82% 10,11% 11,83% 8,73% 9,67% 10,30%
1995 7,79% 7,75% 10,68% 6,42% 7,65% 9,67%
1996 11,76% 9,69% 12,69% 9,60% 10,61% 13,34%
1997 18,10% 12,43% 20,57% 12,28% 15,17% 18,90%
1998 20,25% 13,45% 21,65% 12,80% 18,54% 19,83%
1999 18,04% 14,20% 22,33% 12,44% 14,59% 17,31%
2000 22,18% 19,13% 35,43% 16,87% 18,60% 21,58%
2001 21,52% 17,86% 18,32% 13,28% 20,98% 22,73%
2002 25,97% 25,54% 21,66% 18,91% 26,68% 23,93%
2003 17,03% 15,84% 19,26% 12,21% 18,05% 16,09%
2004 11,07% 9,84% 14,46% 8,78% 12,45% 10,58%
2005 10,24% 10,57% 14,25% 9,43% 11,21% 10,49%
2006 10,02% 8,79% 14,49% 7,77% 10,95% 11,18%
2007 15,96% 17,80% 18,73% 14,22% 16,20% 15,71%
2008 40,89% 49,14% 39,77% 29,65% 43,02% 41,00%
2009 27,23% 43,99% 28,67% 16,83% 34,45% 25,91%
2010 18,01% 16,47% 22,83% 12,71% 20,27% 16,04%
2011 23,23% 25,51% 23,65% 16,37% 25,33% 22,30%
2012 12,74% 11,18% 13,12% 8,34% 14,78% 13,75%
2013 11,04% 11,26% 13,51% 10,22% 12,22% 11,12%
2014 11,35% 10,14% 13,89% 9,21% 11,10% 11,56%
2015 15,46% 16,42% 15,58% 13,54% 15,30% 15,43%
2016 13,07% 17,59% 12,84% 10,99% 15,26% 13,51%
Table 5: Yearly annualized volatility for the smart beta portfolios on the S&P 500 from 1994 to 2016. S&P 500 included for
comparison. Worst (red) and best (green) values for each year are highlighted.

53



Excess Returns vs. S&P 500

Year Value Momentum Low Vol Size Quality
1994 -6,93% 0,66% -4,21% -0,05% 9,28%
1995 -7,76% 0,40% -1,85% -4,53% 4,34%
1996 -7,87% 0,63% -6,55% -1,21% 4,90%
1997 -5,15% -1,94% -4,28% -7,73% 2,69%
1998 -15,14% -2,90% -19,59% -11,88% 6,57%
1999 -27,72% 23,67% -30,21% -11,00% 1,14%
2000 30,29% 33,88% 26,53% 16,86% 11,69%
2001 34,62% 1,93% 13,52% 13,33% 9,82%
2002 13,27% 5,63% 17,63% 7,85% 10,95%
2003 -0,80% 21,08% -6,16% 7,02% -1,73%
2004 4,43% 8,70% 4,01% 6,54% -0,44%
2005 -0,71% 12,94% -2,89% 5,33% 0,22%
2006 6,30% -4,06% 1,60% -0,86% 0,17%
2007 -10,86% 4,53% -4,64% -4,17% 8,24%
2008 5,10% -8,84% 16,38% -2,34% 3,73%
2009 17,20% 0,10% -8,17% 13,83% 7,30%
2010 2,66% -1,40% -3,86% 7,65% -0,19%
2011 0,80% -2,78% 8,00% -0,86% 4,96%
2012 -3,68% -4,59% -6,14% 2,01% -3,04%
2013 -6,30% 5,63% -6,99% 2,07% 1,26%
2014 -0,63% -1,10% 3,92% 3,02% 1,47%
2015 -15,95% 4,04% 2,74% -2,93% -2,56%
2016 19,38% -6,76% -0,85% 4,76% -1,77%
Table 6: Yearly excess returns for the smart beta portfolios on the S&P 500 from 1994 to 2016. S&P 500 included for
comparison. Worst (red) and best (green) values for each year are highlighted.

54



b) Stoxx 600

Annual returns – Smart Beta portfolios vs. Stoxx 600

Year Stoxx 600 Value Momentum Low Volatility Size Quality
2002 -34,53% -4,58% -15,91% -9,36% -31,78% -28,31%
2003 14,22% 27,00% 27,16% 17,49% 27,53% 22,99%
2004 9,76% 19,41% 21,88% 19,50% 14,90% 14,60%
2005 21,55% 17,38% 22,63% 18,72% 22,59% 20,36%
2006 17,20% 16,75% 29,56% 22,97% 22,21% 19,86%
2007 1,12% -16,56% 5,59% -0,49% -0,94% 2,90%
2008 -54,11% -71,03% -44,40% -28,87% -55,39% -40,47%
2009 27,61% 51,18% 14,62% 10,69% 40,70% 38,81%
2010 9,97% 8,81% 27,33% 4,66% 12,31% 12,74%
2011 -9,53% -15,87% -11,76% -1,46% -15,55% -6,54%
2012 14,63% 19,29% 13,91% 13,90% 19,10% 16,41%
2013 16,74% 16,82% 27,82% 15,92% 23,18% 20,24%
2014 5,14% 1,23% 3,83% 10,25% 4,32% 2,11%
2015 8,53% -6,63% 13,21% 11,16% 8,76% 6,75%
2016 0,78% 19,36% 5,87% -2,58% 6,21% 0,33%
Table 7: Yearly annual returns for the smart beta portfolios on the Stoxx 600 from 2002 to 2016. Stoxx 600 included for
comparison. Worst (red) and best (green) values for each year are highlighted.

Sharpe ratios - Smart Beta portfolios vs. Stoxx 600

Year Stoxx 600 Value Momentum Low Volatility Size Quality
2002 -1,29 -0,34 -1,34 -1,15 -2,12 -1,46
2003 0,63 1,94 2,47 2,88 2,01 1,60
2004 0,74 2,56 2,09 3,87 1,90 1,52
2005 1,97 1,74 2,05 2,57 3,51 2,11
2006 0,98 1,06 1,57 2,36 2,17 1,31
2007 -0,21 -1,11 0,07 -0,45 -0,52 -0,10
2008 -1,52 -1,49 -1,59 -1,24 -2,46 -1,23
2009 1,14 1,72 0,68 0,96 2,09 2,09
2010 0,54 0,64 1,61 0,46 0,97 0,98
2011 -0,44 -0,78 -0,63 -0,12 -1,15 -0,36
2012 0,98 1,17 1,14 1,73 1,73 1,33
2013 1,39 1,44 2,16 2,05 2,57 2,06
2014 0,38 0,10 0,26 1,17 0,51 0,19
2015 0,43 -0,37 0,81 0,89 0,79 0,45
2016 0,02 0,85 0,41 -0,23 0,44 0,01
Table 8: Yearly sharpe ratios for the smart beta portfolios on the Stoxx 600 from 2002 to 2016. Stoxx 600 included for
comparison. Worst (red) and best (green) values for each year are highlighted.

55



Annualized volatility – Smart Beta portfolios vs. Stoxx 600

Year Stoxx 600 Value Momentum Low Volatility Size Quality
2002 28,13% 18,39% 13,13% 9,59% 15,76% 20,54%
2003 20,89% 13,39% 10,60% 5,72% 13,17% 13,70%
2004 11,33% 7,03% 9,82% 4,68% 7,11% 8,70%
2005 9,30% 8,14% 9,47% 6,03% 5,53% 8,15%
2006 12,65% 11,19% 15,77% 7,68% 7,99% 11,46%
2007 15,87% 18,88% 17,44% 10,95% 10,37% 15,88%
2008 36,40% 48,55% 28,81% 24,31% 23,06% 33,96%
2009 24,10% 29,69% 21,37% 10,94% 19,38% 18,54%
2010 18,13% 13,57% 16,88% 9,75% 12,62% 12,92%
2011 21,99% 20,36% 18,60% 12,91% 13,60% 18,16%
2012 14,79% 16,47% 12,12% 8,00% 11,00% 12,30%
2013 12,04% 11,62% 12,84% 7,75% 8,99% 9,83%
2014 13,32% 11,85% 14,42% 8,73% 8,47% 10,83%
2015 19,76% 17,78% 16,28% 12,51% 11,08% 14,94%
2016 19,59% 22,36% 13,67% 12,56% 13,54% 14,68%
Table 9: Yearly annualized volatility for the smart beta portfolios on the Stoxx 600 from 2002 to 2016. Stoxx 600 included for
comparison. Worst (red) and best (green) values for each year are highlighted.

Excess returns vs. Stoxx 600

Year Value Momentum Low Volatility Size Quality
2002 29,95% 18,62% 25,17% 2,75% 6,22%
2003 12,78% 12,94% 3,27% 13,31% 8,77%
2004 9,65% 12,12% 9,74% 5,14% 4,84%
2005 -4,17% 1,08% -2,83% 1,04% -1,19%
2006 -0,45% 12,36% 5,77% 5,01% 2,66%
2007 -17,68% 4,47% -1,61% -2,06% 1,78%
2008 -16,92% 9,71% 25,24% -1,28% 13,64%
2009 23,57% -12,99% -16,92% 13,09% 11,20%
2010 -1,16% 17,36% -5,31% 2,34% 2,77%
2011 -6,34% -2,23% 8,07% -6,02% 2,99%
2012 4,66% -0,72% -0,73% 4,47% 1,78%
2013 0,08% 11,08% -0,82% 6,44% 3,50%
2014 -3,91% -1,31% 5,11% -0,82% -3,03%
2015 -15,16% 4,68% 2,63% 0,23% -1,78%
2016 18,58% 5,09% -3,36% 5,43% -0,45%
Table 10: Yearly excess returns for the smart beta portfolios on the Stoxx 600 from 2002 to 2016. Stoxx 600 included for
comparison. Worst (red) and best (green) values for each year is highlighted.

56



Appendix D
Value
a) S&P 500

Daily return annualized volatility

50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

S&P 500 Value

Figure 17: Daily return annualized volatility for the S&P 500 and the Value factor.

a) Stoxx 600

Daily return annualized volatility

50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Stoxx 600 Value

Figure 18: Daily return annualized volatility between 2002 and 2016 for the Stoxx 600 and the Value factor.

57



Momentum
a) S&P 500

Daily return annualized volatility

45%

40%

35%

30%

25%

20%

15%

10%

5%

0%
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

S&P 500 Momentum

Figure 17: Daily return annualized volatility for the S&P 500 and the Momentum factor.

b) Stoxx 600

Daily return annualized volatility

40%
35%
30%
25%
20%
15%
10%
5%
0%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Stoxx 600 Momentum

Figure 20: Daily return annualized volatility between 2002 and 2016 for the Stoxx 600 and the Momentum factor.

58



Low Volatility
a) S&P 500

Daily return annualized volatility

45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

S&P 500 Low Volatility

Figure 21: Daily return annualized volatility for the S&P 500 and the Low Volatility factor.

b) Stoxx 600

Daily return annualized volatility

40%
35%
30%
25%
20%
15%
10%
5%
0%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Stoxx 600 Low Volatility

Figure 22: Daily return annualized volatility between 2002 and 2016 for the Stoxx 600 and the Low Volatility factor.

59



Size
a) S&P 500

Daily return annualized volatility

50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

S&P 500 Size

Figure 23: Daily return annualized volatility for the S&P 500 and the Size factor.

b) Stoxx 600

Annualized daily return volatility

40%

30%

20%

10%

0%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Stoxx 600 Size

Figure 24: Daily return annualized volatility between 2002 and 2016 for the Stoxx 600 and the Size factor.

60



Quality
a) S&P 500

Daily return annualized volatility

45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

S&P 500 Quality

Figure 25: Daily return annualized volatility for the S&P 500 and the Quality factor.

b) Stoxx 600

Daily return annualized volatility

40%

35%

30%

25%

20%

15%

10%

5%

0%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Stoxx 600 Quality

Figure 26: Daily return annualized volatility between 2002 and 2016 for the Stoxx 600 and the Quality factor.

61



Appendix E
Risk free rate of return

U.S. 3-Month Treasury Bill

7%

6%

5%

4%

3%

2%

1%

0%

Figure 27: The U.S. 3-Month Treasury Bill between 1994 to 2016. The Treasury Bill is referred to as the risk free rate.

62