DEGREE PROJECT IN INDUSTRIAL ENGINEERING AND MANAGEMENT, SECOND CYCLE, 30 CREDITS STOCKHOLM, SWEDEN 2019
A Three-Pronged Sustainability-Oriented Markowitz Model Disruption in the fund selection process?
SIMON LOUIVION
EDWARD SIKORSKI
KTH ROYAL INSTITUTE OF TECHNOLOGY SCHOOL OF INDUSTRIAL ENGINEERING AND MANAGEMENT Abstract
Since the term ESG was coined in 2005, the growth of sustainable investments has outpaced the overall asset management industry. A lot of research has been done with regards to the link between sustainability and financial performance, despite the fact that there is a lack of transparency in sustainability of listed companies. This thesis breaks down the word sustainability into two di↵erent categories, and in turn eleven di↵erent parameters. The result is the term Q score which represents a company’s sustainability. The purpose is to increase transparency in the fund selection process for asset managers. Further, a multi- objective optimization problem is solved to analyze the relationships between return, risk and sustainability. The main subject is that accommodating sus- tainability as a third parameter in addition to return and risk modifies the fund selection process. The result indicates that the relationships between sustain- ability, return and risk follow the e cient market hypothesis, implying that an investor would have to sacrifice risk and return in order to achieve higher sus- tainability. With that said, the results indicated that the sacrifice is relatively small, and that there are a number of sustainable portfolios that perform well. Moving on, the reporting of ESG company data is still lacking. For this reason, this master thesis acts as a precursor for any future development within the field.
Key words: Portfolio optimization, e cient frontier, multi-objective optimiza- tion problem, sustainability, ESG, e cient market hypothesis, behavioural fi- nance En tre-dimensionell h˚allbarhetsorienterad Markowitz modell
Sammanfattning
Sedan termen ESG utvecklades ˚ar2005, har tillv¨axten av h˚allbara investeringar vuxit snabbare ¨anden generella f¨orvaltningsindustrin. Mycket forskning har gjorts kring h˚allbarhet kopplat till finansiell avkastning, men trots detta saknas det fortfarande en transparens r˚adande h˚allbarhet av noterade bolag. Detta examensarbete bryter ned termen h˚allbarhet till tv˚akategorier, vilket i sin tur bryts ner till elva kvantifierbara parametrar. Resultatet blir ett s˚akallat Q score, som ¨arett v¨arde p˚aett f¨oretags h˚allbarhet. Syftet med arbetet ¨aratt ¨oka transparensen av fonders h˚allbarhetsarbete. Vidare l¨oses ett optimeringsprob- lem med tre parametrar f¨oratt unders¨oka f¨orh˚allandena mellan avkastning, risk och h˚allbarhet. Resultatet indikerar att dessa f¨orh˚allanden f¨oljer hypotesen om e↵ektiva marknader, vilket inneb¨aratt en investerare m˚aste o↵ra avkastning och risk f¨oratt uppn˚aen mer h˚allbar portf¨olj. Med det sagt, indikererar resul- tatet att en investerare inte beh¨over o↵ra mycket inom avkastning f¨oratt uppn˚a en h˚allbar portf¨olj. Vidare kvarst˚ardet mycket arbete inom rapporteringen av ESG data p˚af¨oretagsniv˚a.Av detta sk¨alanses detta examensarbete vara en f¨oreg˚angare innan datan utvecklas vidare.
Nyckelord: Portf¨oljoptimering, e cient frontier, portf¨oljteori, h˚allbarhet, ESG, hypotesen om e↵ektiva marknader, beteendeekonomi Preface
This Master’s thesis was written in the spring of 2019 by Simon Louivion and Edward Sikorski, during their five-year master’s degree program within Indus- trial Engineering and Management at KTH Royal Institute of Technology.
We would like to take this opportunity to thank our supervisors Sofia W¨arml¨of Helmrich and Christian Thomann at SEB and KTH respectively. We are also thankful to Fredrik Armerin for his guidance with the optimization problem. Lastly, a final thank you to our parents for their never-withering belief in us. We would not be here without them. ”Earth provides enough to satisfy every man’s needs, but not every man’s greed.”
MAHATMA GANDHI
4 Contents
1 Background 1 1.1 Aim ...... 2 1.2 Scope ...... 2 1.3 Research question and problem statement ...... 2 1.4 Expectedcontribution ...... 3
2 Literature Review and Theory 5 2.1 Definitions...... 5 2.1.1 ESG...... 5 2.1.2 Impact ...... 6 2.1.3 Screeningmethods ...... 6 2.2 Modern portfolio theory ...... 6 2.2.1 Multi- optimization problem ...... 6 2.2.2 Markowitz’s mean-variance model ...... 7 2.2.3 Solution to the mean-variance problem ...... 7 2.2.4 The e cientfrontier ...... 8 2.2.5 Trade-o↵problem ...... 9 2.3 E cientmarkethypothesis ...... 9 2.4 Behavioural finance ...... 10 2.4.1 Herding ...... 10 2.4.2 Availability ...... 11 2.4.3 Home bias ...... 11 2.4.4 Investing in sustainability ...... 11 2.4.5 Overconfidence ...... 11 2.4.6 Present bias or short-termism ...... 12 2.4.7 OthercriticismofEMH ...... 12 2.5 Empiricalresearch ...... 13 2.5.1 Sustainability and performance ...... 13 2.5.2 Sustainability and risk ...... 14 2.5.3 Theory versus practice ...... 14 2.5.4 Real-world impact ...... 15 2.5.5 Three-criterion Markowitz models ...... 16 2.5.6 Morningstar and Sustainalytics’ ESG criteria ...... 16
3 Method 19 3.1 Literaturestudy ...... 19 3.2 Data collection ...... 19 3.2.1 Static company data ...... 19 3.2.2 Financial company data ...... 20 3.2.3 Sustainability company data ...... 20 3.2.4 Investmentfundsdata ...... 21 3.3 IMTdevelopment...... 21 3.3.1 Data transformation ...... 21 3.3.2 ESG parameters ...... 22 3.3.3 EquityESGscore ...... 22 3.3.4 Impact parameters ...... 23 3.3.5 EquityQscore ...... 23 3.3.6 Fund Q score ...... 23 3.3.7 Fund ratings ...... 24 3.4 Optimization problem ...... 24 3.4.1 Calculating returns and covariance from historical data . 25 3.4.2 A sustainability-oriented Markowitz model ...... 25
4 Results 27 4.1 ESGfactors...... 27 4.1.1 Carbon intensity ...... 27 4.1.2 Wasteintensity...... 28 4.1.3 Waterstressintensity ...... 28 4.1.4 Diversity ...... 30 4.1.5 Labour conditions ...... 31 4.1.6 Corporate governance ...... 31 4.2 Impact factors ...... 32 4.2.1 Job creation ...... 32 4.2.2 Other impact indicators ...... 33 4.3 IMTQscores...... 35 4.3.1 ESG scores ...... 35 4.4 Countryscores ...... 36 4.5 Industryscores ...... 37 4.6 Sectorscores ...... 38 4.7 Fundanalysis...... 39 4.7.1 Di↵erence between long-only and long-short funds . . . . 40 4.7.2 Non-sustainable versus sustainable funds ...... 40 4.7.3 Positive versus negative screening ...... 40 4.8 Optimization solution ...... 41
5 Discussion 46 5.1 IMTmodel ...... 46 5.1.1 Model validation ...... 47 5.2 Sustainability-oriented Markowitz model ...... 48 5.3 Generaldiscussion ...... 50
6 Conclusion 53 6.1 Answering the research questions ...... 53 6.2 Implications...... 54 6.3 Furtherresearch ...... 55
References 56
7 Appendices 61 7.1 Graphs and tables ...... 61 List of Abbreviations and Glossary
Alpha - A term used in investing describing a strategy’s ability to outperform the market.
CSR - Corporate social responsibility
EBIT - Earnings before interest and taxes IMT - Impact metric tool
Maximum drawdown - maximum loss from a peak to a trough of a portfo- lio
MPT - Modern portfolio theory
NAV - Net asset value
SDG - Sustainable development goals, set by the United Nations
SRI - Socially responsible investing
UN - United Nations
7 1 Background
The world is changing, for the better and for the worse. Within the asset man- agement industry, sustainable investing has become a growing thematic. The term ESG (Environmental, Social and Governance) has grown significantly since it was first coined in 2005 (Hagart and Knoepfel 2005). In the past decade, the growth of sustainable investments has significantly outpaced the overall asset management industry. Investors have been pressured to adapt to society’s es- calated demand for sustainable investing. More than one in four dollars under professional management in the US was invested in Socially Responsible In- vesting (”SRI”) strategies by the end of 2018 (The Forum for Sustainable Re- sponsible Investment 2018). In Europe, 50% of assets were invested in Socially Responsible Investing (”SRI”) strategies at the end of 2015. (Global Sustain- able Investment Alliance 2016)
The salience of sustainable investing is partly due to pressure from asset own- ers. It is quite common that asset managers have to meet a certain ”ESG-level”, defined by the asset management firm and its clients. A commonly used mea- surement of ESG, is Morningstar’s sustainability ratings. In March of 2016, they first published their ESG ratings of more than 20 000 mutual funds. This was certainly groundbreaking, as investors had previously no simple way of mea- suring the sustainability of mutual funds. A study published in 2018, examined that there was an outflow of $12 billion from low sustainability funds, while high sustainability categorized funds had net inflows of $24 billion since the introduction of Morningstar’s ESG fund ratings. The multi-billion dollar move- ment of funds occurred during an eleven month period after the publishing of Morningstar’s sustainability ratings, proving through causality the tremendous interest for sustainable investing. This does, however, beg the question whether the financial markets have all the information available with regards to sustain- ability. A $24 billion shift as a result of one company’s actions proves that there was a lack of knowledge, and that perhaps that the absence still remains. (Hartzmark and Sussman 2018)
The purpose of this master thesis is to develop a universal framework that can be used to measure the overall sustainability of any portfolio of listed com- panies. This will aid in providing investors with additional information in the fund selection process, whilst also introducing new dimensions when construct- ing portfolios. Lastly, as more and more asset managers are required to allocate a certain amount of funds into sustainable investment as a fiduciary duty; it is of interest to see how they allocate those funds and how an increase in sustain- ability will a↵ect the risk and return of their portfolios.
The relationship between risk and return has been studied extensively since Markowitz laid out the foundation of Modern Portfolio Theory in 1952. During the past 67 years since then, there has been a tremendous change in the finan- cial industry. Technology, risk appetite and other factors of importance have
1 altered investor preferences. Markowitz’s MPT is a theory on how investors can construct portfolios to maximize their expected return, given a specific level of risk. (Markowitz 1952) As a result of the salience in demand for sustainability, we intend to extend Markowitz’s optimization problem with a third criterion. The third criterion is a sustainability measure developed throughout this thesis, which is used and tested in the optimization problem. Additionally, previous re- search has commonly performed similar optimizations with long-only portfolios, which we refer to in section 2. We intend to add to research by also examining long-short funds and how they meet their ESG criterion. Previous research has also performed the analysis using di↵erent forms of utility functions, which will also be covered in this thesis. There has also been done extensive research using regression tools, which is an excellent tool to examine relationships between pa- rameters. However, by using an optimization problem instead, the relationships are easier to apprehend while contributing to this avenue of research.
1.1 Aim The purpose of this master thesis is to develop a universal, factor agnostic frame- work for SEB. The framework can be modified to incorporate any investor’s intentions. The aim can be broken down into three parts. Provide investors with additional information in the fund selection process • Introduce new dimensions when constructing portfolios • Evaluate investment policies in relation to sustainability goals • 1.2 Scope The developed framework must be universal and include every listed company worldwide. The framework has to be able to analyze the sustainability of any given fund. Our data set covers both long and long-short equity funds as well as the equities listed on OMXS30. We will also use SEB’s previously de- fined parameters in the framework. Moreover, we will build a three-parameter Markowitz model in order to calculate how funds and equities perform with respect to risk and return, in relation to their sustainability level. This devel- oped framework will hopefully aid SEB in their client meetings, giving investors further transparency.
1.3 Research question and problem statement With the thematic of sustainable investing undergoing rapid growth, investors are becoming pressed by asset owners to meet certain criteria of sustainability. They are also pressed into knowing specifically what their investments are im- pacting. However, this information in today’s financial industry is either scarce or expensive. General ESG scores and CSR indices are too general to analyze, as they do not break down specifically what a company is exactly doing to provide
2 real-world impact. What does a good ESG rating actually signify? What is the company good at? These are questions that remain unanswered by a general ESG score. A framework, that is fully transparent in its parameters, will be developed in this master thesis.
It is di cult to measure sustainability as there are many angles to a company’s sustainability. A company that is manufacturing solar panels may have a pos- itive impact. However, the company may also be linked to treacherous labour conditions and poor corporate governance. In order for financial institutions to be able to present investors a transparent holistic perspective of any fund, the sustainability factor of a fund has to be broken down to di↵erent parameters. These parameters have to be quantifiable and normalized for comparison. Given these parameters, will that alter the fund selection process? Previous research has stated that investors are more likely to avoid ”sinful stocks” than to invest in virtuous stocks (Johansson 2019).
The research question culminates in the following two problem statements:
Can a framework quantify the sustainability of funds? • How will the implementation of a diverse sustainability measure impact • portfolios in relation to risk and return, and how may it a↵ect investors in their decision making?
The developed framework will be called the Impact Metric Tool (IMT) and will be used by SEB to assess the sustainability of investment funds.
1.4 Expected contribution Previous research in sustainability-oriented Markowitz models has increased during the last five years. As mentioned previously, we aim to add to this field of research by expanding the third criterion. The third criterion is of im- portance, as Utz et al. (2015) concluded that there is a need for increased service in relation to investing in sustainable funds. We believe that the quantification of sustainable factors will increase transparency for investors, allowing them to precisely pinpoint what their investments are contributing to. The framework should not be a single measurement, but rather show what each parameter has contributed to the overall rating of each equity/fund.
Previous research examining the relationship between stock price performance and sustainability has been somewhat inconclusive, which we delve into in sec- tion 2.5.1. Our thesis will contribute to the examining of how long-short funds meet their sustainability targets. Their ability to short an equity, gives a fur- ther dimension to the existing research. If a fund goes short a company with a negative sustainability score, this will result in a positive contribution to the portfolio score. Thus, our optimization problem intends to examine the rela- tionship between sustainability, risk and return given the ability to be able to
3 go short an equity. In addition, research on financial performance and sustain- ability has been extensively conducted using regression tools. As mentioned, this thesis will instead use a Markowitz model to analyze the relationships.
4 2 Literature Review and Theory
In this section, definitions will first be set to mitigate any potential confusion. Subsequently, relevant literature and theory will be presented within all fields of the master thesis. Modern portfolio theory, the e cient market hypothesis and behavioural finance are the three main pillars of this chapter.
2.1 Definitions Parallel to the emergence of sustainability-oriented funds and investment strate- gies, there has been a plethora of terms used to describe sustainability. We intend to distinguish the di↵erence between di↵erent terms in order to leave no room for confusion. These definitions have been set in collaboration with SEB, as well as from the white paper written in preparation for the thesis. (Johansson 2019)
Sustainability is the most used word in the space of Socially Responsible In- vesting strategies for obvious reasons. The word does, however, lack a certain dimension to it. For example, the exclusion of controversial stocks such as oil stocks is defined as sustainable investing. However, does the exclusion of the world’s largest oil producer ExxonMobil really contribute to making the world a better place? In addition, software companies receive in general positive ESG scores, due to their relatively low carbon emissions. However, one might question if these companies are producing positive real-world impact. For this reason, this thesis divides the term sustainability into two subcategories, namely ESG and Impact.
2.1.1 ESG The term ESG is today widely used in the financial industry to describe the environmental, social and governance aspects of a company. The abbreviation is successful in describing what is of importance for investors, as the three pil- lars are exhaustive in sustainability terms. In this thesis, the term will denote a company’s operational e ciency in relation to externalities. For example, the terms ROI or operating margin can be used as a metric for evaluating the fi- nancial e ciency of a company. In the case of ESG, the waste intensity metric quantifies environmental e ciency, in a similar way.
The concept of ESG intends to entail how a company’s business is conducted. Demonstrating favourable ESG characteristics does not always signify the cre- ation of impact. To exemplify this, a company might be highly e cient in terms of having a low carbon footprint. However, if the company does not try to further improve itself; there is no impact being created. The ESG factors measure the e ciency of a company, but they do not include the improvement of operations.
5 2.1.2 Impact The shortcomings of the term ESG, are aimed to be covered by introducing the term impact. Using the Sustainable Development Goals (SDGs) developed by the United Nations, we intend to define the term impact. We use these goals to measure whether a company is having a real-world impact. There are a plethora of ways to create an impact in the world. Most importantly, companies that engage in the production of goods or services that are aligned with any of the SDGs have an impact solely based on their output. Other ways a company can have an impact is through their operations. If a firm has employees in devel- oping countries, this could align with the SDG ”No poverty” as job creation counteracts poverty. In essence, a company can both directly provide impact with its products, but also indirectly through how the company conducts its business. (United Nations 2015)
It is of importance now to distinguish the di↵erence between impact and ESG, as impact refers to what a company does, as opposed to ESG capturing how a company does it.
2.1.3 Screening methods Positive screening implies a screening method with the inclusion of sustain- able companies. Examples of industries that these companies would operate in are waste management, public transport, education, environmental technology and renewable energy.
Negative screening revolves around the exclusion of ”bad” companies. This methodology can either be done by excluding entire sectors. These industries include but are not limited to oil & gas, tobacco, mining and pornography. An- other common approach is by setting a minimum ESG score for a security in the screening process. If a company has an ESG score below a certain level, it would be excluded from the screening process. This approach can be argued as non-sustainable, as the fund does not necessarily invest in companies working towards a more sustainable society. However, in this thesis, funds are considered sustainable as long as they apply a screening method, regardless if positive or negative.
2.2 Modern portfolio theory This chapter will include the mathematical theory used in the thesis. Markowitz (1952), Armerin (2004), Zopounidis et al. (2015) and Hult et al. (2012) were used as sources of literature.
2.2.1 Multi- optimization problem Multi-objective optimization is a commonly used methodology to solve portfolio selection problems. It is a method for dealing with a mathematical problem
6 where than objective function has to be optimized over more than one constraint. In some cases, these objective functions are conflicting, which allows for multiple solutions. These problems are defined as follows:
optimize F (x)=[f1(x),...,fk(x)] (1) subject to x X 2 x is a vector containing decision variables while X is a set of feasible solutions. F (x) is regarded as the objective function, while x X is regarded as the constraints. The objective function can either be minimized,2 maximized or both. The purpose of multi-objective optimization is to find all optimal solutions. (Zopounidis et al. 2015)
2.2.2 Markowitz’s mean-variance model Modern Portfolio Theory (MPT), first introduced by Markowitz in 1952, is based on a portfolio optimization problem with two criteria: expected return and risk. In the model, risk is measured by using variance to quantify the variability of the return. Expected return may be calculated using random variables, but it can also be proxied by using historical returns. (Markowitz 1952) The latter will be the case in this thesis.
MPT states that an investor can construct a portfolio of multiple assets for a certain level of risk. A key assumption to MPT is that investors are risk- averse, i.e. they prefer to invest in a less risky portfolio than a riskier portfolio for a certain level of return. In turn, this implies that investors are willing to take on more risk, in order to achieve higher returns. (Markowitz 1952)
The mean-variance model can be formulated as an optimization problem, with the objective function to minimize the portfolio’s expected risk, with the con- straint of attaining a certain level of return. 1 minimize wT ⌃w 2 subject to µT w µ (2) min wT 1 =1 w is a vector with the weight of each asset. ⌃is the covariance matrix of the assets and µ is a vector with the expected return of each asset. µmin is the minimum level of return the investor wishes to obtain. (Armerin 2004)
2.2.3 Solution to the mean-variance problem The mean-variance problem introduced in the previous section can be solved by using the Lagrange multiplier method (Markowitz 1952). In this problem, the Lagrangian function is written as 1 (w, , )= wT ⌃w + (µ wT µ)+ (1 wT 1) (3) L 1 2 2 1 min 2
7 Following this, the Lagrange multiplier method states: