Embed Size (px)
Citation preview
Escola de Economia e Gestão
Orlanda Cristina Araújo Baptista
The impact of ESG criteria on
portfolio financial performance
Outubro de 2019
Universidade do Minho
Escola de Economia e Gestão
Orlanda Cristina Araújo Baptista
The impact of ESG criteria on portfolio
financial performance
Dissertação de Mestrado
Mestrado em Finanças
Trabalho efetuado sob a orientação do(a)
Professora Doutora Benilde Maria do
Nascimento Oliveira
Outubro de 2019
ii
DIREITOS DE AUTOR E CONDIÇÕES DE UTILIZAÇÃO DO TRABALHO POR TERCEIROS
Este é um trabalho académico que pode ser utilizado por terceiros desde que
respeitadas as regras e boas práticas internacionalmente aceites, no que concerne aos
direitos de autor e direitos conexos.
Assim, o presente trabalho pode ser utilizado nos termos previstos na licença abaixo
indicada.
Caso o utilizador necessite de permissão para poder fazer um uso do trabalho em
condições não previstas no licenciamento indicado, deverá contactar o autor, através
do RepositóriUM da Universidade do Minho.
Licença concedida aos utilizadores deste trabalho
Atribuição-NãoComercial-CompartilhaIgual CC BY-NC-SA
https://creativecommons.org/licenses/by-nc-sa/4.0/
Universidade do Minho, ___/___/______ Assinatura: ________________________________________________
iii
Acknowledgements
I am proud for reaching this moment and grateful to all the people that support
me throughout this process.
I would like to sincerely express my gratitude to my supervisor, Professor Doctor
Benilde Oliveira, for her guidance, willingness in clarifying my doubts, all her effort and
availability to help me.
Additionally, I would like to acknowledge Professor Doctor Gilberto Loureiro’s
help and availability, in guiding me through the software used on my dissertation.
I would like to thank all my friends from Madeira, that joined me on this journey
in University of Minho. They were like a family to me and made the distance from home
feel shorter.
I want to dedicate this dissertation to my parents that have always dreamt this
journey for me. I am lucky and truly happy for having you always by my side, your
unconditionally support and love.
To the most important person in this journey, my boyfriend Marco Santos, that
made me get out of my comfort zone, be more ambitious and dream bigger. For your
patient, love, encouragement and especially emotional support, I want to thank you. I
will be grateful for the rest of my live for having you during this challenging path.
iv
STATEMENT OF INTEGRITY
I hereby declare having conducted this academic work with integrity. I confirm that I
have not used plagiarism or any form of undue use of information or falsification of
results along the process leading to its elaboration.
I further declare that I have fully acknowledged the Code of Ethical Conduct of the
University of Minho.
Universidade do Minho, ___/___/______ Assinatura: _______________________________________________
v
The impact of ESG criteria on portfolio financial performance
Abstract
The increasing demand of investors for SRI, in the last 30 years, has stimulated
the scientific community to study the impact of ESG criteria on investment financial
performance. Theory suggests that if it was beneficial to invest in SRI, this advantage
disappeared as soon as markets participants fully incorporated the information about
SRI firms. More recent studies state that investors not only hold SRI because of their
ethical beliefs but also because it reduces their downside risk. The aim of this
dissertation is to analyse if different ESG dimensions impact SRI financial performance
during different market conditions. For this purpose, it is constructed and assessed the
performance of two distinct portfolios based on ASSET4 ESG scores. The monthly sample
comprises US SRI companies from 2002 to 2017. Portfolio performance is assessed on
the basis of the popular Carhart (1997) four-factor model and the recent Fama and
French (2015) five-factor model. Several robustness checks, for alternative weighted
scheme, screen approach, different cut-offs and the exclusion of financial firms, are
implemented. Additionally, a dummy constructed based on NBER business cycles to
account for different market conditions, is added to the models. The results suggest that
if investors implement a Long-Short strategy based on the GOV or ESG dimensions (and
SOC in the case of the five factor-model), they obtain negative abnormal returns. The
results for earlier sub-periods show that ESG portfolios have, in fact, a neutral financial
performance, however, the tendency seems to have changed in 2012 to a negative
financial performance. Moreover, by including a dummy variable in the models it is
shown that portfolio performance does not significantly change according to the state
of the economy.
Keywords: ESG criteria; Performance of Stock Portfolio; Recession Periods; Socially
Responsible Investments; Time-varying Performance
vi
O impacto das dimensões ESG na performance financeira dos portfólios
Resumo
A crescente procura dos investidores pelos ISR, nos últimos 30 anos, estimulou a
comunidade científica a estudar o impacto dos critérios de ESG na performance
financeira do investimento. A teoria sugere que era benéfico investir em ISR mas esta
vantagem desapareceu assim que os participantes do mercado incorporaram toda a
informação sobre as empresas SR. Estudos mais recentes referem que os investidores
não só detêm ISR por causa das suas crenças éticas, mas também porque estes
investimentos são capazes de reduzir o risco potencial negativo. O objetivo desta
dissertação é analisar se diferentes dimensões de ESG impactam a performance
financeira dos ISR durante diferentes condições de mercado. Para este propósito, foram
construídos e medida a performance de dois portfólios distintos com base nos índices
de ESG da ASSET4. A amostra mensal engloba empresas dos EUA entre 2002 e 2017. A
performance é medida através do modelo de quatro fatores de Carhart (1997) e do
recente modelo de cinco fatores do Fama e French (2015). Foram implementados
diversos testes de robustez, como um alternativo esquema de ponderação do portfólio,
abordagem de construção do portfólio, diferentes cut-offs e exclusão de empresas
financeiras. Além disso, uma dummy construída com base nos ciclos de negócio da
NBER, para capturar as condições de mercado, é adicionada aos modelos. Os resultados
sugerem que se os investidores implementarem uma estratégia Long-Short baseada nas
dimensões de GOV ou ESG (e SOC no caso do modelo de cinco fatores), irão obter
retornos negativos. Os resultados para os primeiros subperíodos mostram que, de facto
os portfólios ESG apresentam performance financeira neutra, contudo, a tendência
parece ter mudado em 2012 para uma performance financeira negativa. Incluindo uma
variável dummy nos modelos verifica-se que a performance dos portfólios não se altera
significativamente de acordo com os estados da economia.
Keywords: critérios de ESG; Performance de portfólios de ações; Períodos de Recessão;
Investimentos Socialmente Responsáveis; Performance estado-dependente
vii
Table of Content
Acknowledgements .............................................................................................................................. 3
Abstract ................................................................................................................................................ 5
Resumo ................................................................................................................................................. 6
List of Acronyms ................................................................................................................................... 8
List of Tables ......................................................................................................................................... 9
List of figures ...................................................................................................................................... 10
List of Appendices .............................................................................................................................. 11
1.Introduction..................................................................................................................................... 13
2.Literature Review ............................................................................................................................ 16
2.1 An overview about SRI ............................................................................................................. 16
2.2 The performance of SRI versus the performance of conventional strategies ........................ 17
2.3 Empirical evidence on the performance of SRI ....................................................................... 18
2.4 Investing in SRI: “the learning hypothesis” ............................................................................. 23
2.5 SRI in Recession Periods ........................................................................................................... 24
3.Methodology and Dataset .............................................................................................................. 27
3.1 ESG Portfolio Construction ...................................................................................................... 27
3.2 Performance Measurement ..................................................................................................... 28
3.3 Recession Periods ..................................................................................................................... 29
3.4. Dataset Description ................................................................................................................. 31
4. Results ............................................................................................................................................ 34
4.1 Performance Evaluation ........................................................................................................... 34
4.2 Robustnes checks: weightning scheme, screen approach, different cut-offs and the
exclusion of financial firms ............................................................................................................ 37
4.3 Performance in expansion versus recession periods .............................................................. 42
5. Conclusion ...................................................................................................................................... 46
References .......................................................................................................................................... 48
Appendix............................................................................................................................................. 53
viii
List of Acronyms
Acronyms
ESG
ENV
GOV
NBER
SOC
SRI
US
Description
Environmental, Social and Governance
Environmental
Governance
National Bureau of Economic Research
Social
Socially Responsible Investing/Investments
United States
ix
List of Tables
Table 1- Descriptive statistics of portfolios’ returns ........................................................................... 33
Table 2 - Portfolio performance estimates- Carhart (1997) four-factor model ................................. 35
Table 3- Portfolio performance estimates- Fama and French (2015) five-factor model.................... 36
Table 4 - Long-short portfolio performance estimates depending on the weighting scheme ........... 38
Table 5 - Long-short portfolio performance estimates depending on the screen approach ............. 39
Table 6- Long-short portfolio performance estimates depending on the cut-off .............................. 40
Table 7- Long-short portfolio performance estimates for different subperiods ................................ 41
Table 8- Long-short performance estimates of portfolios with and without financial firms ............. 42
Table 9- Portfolio performance estimates - Carhart (1997) four-factor model with dummies ......... 44
Table 10- Portfolio performance estimates - Fama and French (2015) five-factor model with
dummies ............................................................................................................................................. 45
x
List of figures
Figure 1- The growth evolution of SRI in the United States (source: USSIF 2016 Trends report) ...... 13
Figure 2- Monthly evolution of the US Market Portfolio, available on the Professor Kenneth French
webpage, over the period of January 2002 to September 2017. ....................................................... 30
Figure 3- Monthly excess returns of the US Market portfolio, available on the Professor Kenneth
French webpage, over the period of January 2002 to September 2017. ........................................... 30
xi
List of Appendices
Appendix A - Definition of Market States according to NBER Business cycles ................................... 53
Appendix B - Descriptive statistics of value weighted portfolios returns ........................................... 54
Appendix C – Portfolio performance estimates of the value-weighted scheme– Carhart (1997) four-
factor model………………………………………. ............................................................................................. 54
Appendix D – Portfolio performance estimates of the value-weighted scheme – Fama and French
(2015) five-factor model ..................................................................................................................... 55
Appendix E - Descriptive statistics of Best-in-Class portfolios returns ............................................... 56
Appendix F – Portfolio performance estimates of the Best-in-Class approach- Carhart (1997) four-
factor model…………............................................................................................................................. 56
Appendix G – Portfolio performance estimates of the Best-in-Class approach- Fama and French
(2015) five-factor model ..................................................................................................................... 57
Appendix H – Portfolio performance estimates depending on the cut-off- Carhart (1997) four-factor
model…………….………………………………………………………………………………………………………………………………57
Appendix I - Portfolio performance estimates depending on the cut-off- Fama and French (2015) five-
factor model………………………………………………………………………………………………………………………………….58
Appendix J - Portfolio performance estimates for different subperiods- Carhart (1997) four-factor
model…………………………………………………………………………………………………………………………………………..58
Appendix K - Portfolio performance estimates for different subperiods- Fama and French (2015) five-
factor model………………………………………………………………………………………………………………………………….59
Appendix L - Descriptive statistics of portfolio returns without financial firms ................................. 60
Appendix M - Performance estimates of portfolios without financial firms– Carhart (1997) four-factor
model….…………………………………………………………………………………………………………………………………………60
Appendix N - Performance estimates of portfolios without financial firms– Fama and French (2015)
five-factor model ................................................................................................................................ 61
Appendix O – Portfolio performance estimates of the value weighted scheme – Carhart (1997) four-
factor model with dummies ................................................................................................................ 62
Appendix P - Portfolio performance estimates of the value weighted scheme – Fama and French
(2015) five-factor model with dummies ............................................................................................. 63
Appendix Q - Portfolio performance estimates of the Best-in-Class screen approach – Carhart (1997)
four-factor model with dummies ........................................................................................................ 64
Appendix R - Portfolio performance estimates of the Best-in-Class screen approach – Fama and
French (2015) five-factor model with dummies ................................................................................. 65
Appendix S - Portfolio performance estimates depending on the cut-off– Carhart (1997) four-factor
model with dummies .......................................................................................................................... 66
Appendix T - Portfolio performance estimates depending on the cut-off – Fama and French (2015)
five-factor model with dummies ........................................................................................................ 67
xii
Appendix U – Performance estimates of portfolios without financial firms– Carhart (1997) four-factor
model with dummies .......................................................................................................................... 68
Appendix V – Performance estimates of portfolios without financial firms– Fama and French (2015)
five-factor model with dummies ........................................................................................................ 69
13
1.Introduction
Socially Responsible Investing (SRI) is an investment style that considers the
incorporation of environmental, social and governance (ESG) criteria, with the intent to
obtain competitive financial returns while positively impacting the society. Market
participants engaging in SRI not only want to have financial products consistent with
ethical values, but they also seek to achieve long-term competitive financial returns,
manage risk, fulfil fiduciary duties or even contribute to the development of ESG
practices (USSIF 20171).
Over the last few decades, SRI has experienced an exponential growth around
the world. According to USSIF 2016 Trends report 2 , the total SRI assets under
management grew 33% between 2014 and 2016 and in the beginning of 2016
represented $8.72 trillion. Moreover, the number of investment funds incorporating
ESG criteria grew between 1995 and 2016, from 55 with a total net asset of $12 billion
to 1002 with a total net asset of $2597 billion (figure 1).
Along with the growth of interest in SRI there has been an increase in the volume
of research studying the impact of ESG criteria on investment financial performance.
1 Based on The Forum US Sustainable, Responsible and Impact Investment website – see https://www.ussif.org/sribasics and https://www.ussif.org/esg. 2 According to annual USSIF Report on US sustainable, Responsible and Impact Investing Trends 2016.
Figure 1- The growth evolution of SRI in the United States (source: USSIF 2016 Trends report)
14
However, the existing literature presents puzzling results about SRI performance.
Despite some evidence that SRI delivers positive abnormal returns, there seems not to
be a consensus on which ESG dimension or which type of screen approach (positive,
negative or Best-in-Class) positively impacts SRI financial performance.
In fact, due to the fast growth experienced by SRI, there was information
asymmetries in the 1990s that financially favoured SR investors. Since most events that
shaped SRI occurred in 1980 (Renneboog et al., 2008a and Bebchuk et al., 2013), in the
1990s investors and financial analysts did not have enough skills to perceive the
performance difference between companies with good and bad CSR practices, so, at
that time, SRIs were not yet correctly priced, providing abnormal returns to investors.
However, under the Bebchuk et al. (2013) “learning hypothesis”, market participants
gradually acknowledged the differences between SR and non-SR firms and once a
sufficient number of investors fully incorporated these differences, stocks became
correctly priced and the positive association between ESG criteria and financial
performance disappeared. Additionally, many researchers have captured the
persistence of abnormal returns in the 1990s and the consequent disappearances in the
early 2000s (e.g. Derwall et al., 2011; Bebchuk et al., 2013 and Borgers et al., 2013).
Consequently, researchers have been questioning the reason why SRI demand
continues to increase if they no longer are able to outperform the market. Even if
“values-driven” investors are willing to quit some financial benefits in order to have
investments consistent with their believes, the “profit-seeking” 3 investors do not.
Additionally, Nofsinger and Varma (2014) suggest that investors can reduce their
downside risk because SRIs perform better during recession periods.
The purpose of this dissertation is to analyse the relationship between ESG
criteria and portfolio performance, while controlling for market states. In this context,
using rankings constructed based on Asset4 ESG scores, two portfolios scoring distinct
in ESG criteria are constructed for each ESG dimension. In addition, to analyse the
differential impact of the ESG criteria on financial performance between the two
3 Nilsson (2009) identified three types of SRI investors: those who base their investment decision only on risk/return (“profit-seeking” investor); those who base their investment decision only on Social Responsibility (“values-driven” investor); and finally, those who make their investment decision based simultaneously on return and Social Responsibility.
15
portfolios, a Long-Short strategy is followed. This methodology not only allows the
independent analysis of the impact of each ESG dimension on portfolio performance,
but it also overcomes the limitations related to SR funds’ performance. The labelled SR
funds sometimes do not maintain their SR status over time or invest in stocks with lower
ESG scores than their conventional counterparts (Auer, 2016; Wimmer, 2013 and Henke,
2016).
In order to assess the SR portfolios’ performance, the popular Carhart (1997)
four-factor model and the recent Fama and French (2015) five-factor model are used.
However, these models do not allow risk and return to vary over time, which can provide
biased results. Some recent studies (e.g. Silva and Cortez, 2016; Henke, 2016 and
Nofsinger and Varma, 2014) argue that the performance of SRI may be state dependent,
and these studies therefore advocate the use of models that allow for risk and return to
vary according to different market states. Consequently, in this study a dummy variable
is added to the four and five-factor models to allow for risk and return to vary according
to the NBER business cycles of recession and expansion.
To the best of my knowledge, this dissertation contributes to the existing
literature in the extent that it assesses synthetic portfolio performance during different
market conditions and also offers the perception of how each ESG dimension impacts
portfolio performance. Additionally, this study gives more recent insight about US SRI
performance. The sample is comprised of 2357 US companies from January 2002 to
September 2017. This sample is larger than the one used by Halbritter and Dorfleitner
(2015) (concerning Asset4 data) and allows a sight to whether or not SRI financial
performance has persisted over the last years.
The remainder of this dissertation is developed into 4 additional sections. The
following section (section 2) summarizes and discusses the most relevant studies
concerning SRI and portfolio performance. In section 3, the dataset is described, the
definition of market states is presented, and also the methodology implemented to
assess portfolio performance is described. Section 4 reports and discusses the results.
Finally, section 5 concludes and presents some limitations of this dissertation.
16
2.Literature Review
This section reviews, summarizes and discusses the most relevant studies on SRI.
Firstly, it gives an overview about SRI and its popularity, and afterwards, a discussion on
the performance of SRI as well as on the methodologies used to assess it, is presented.
The last part of this section is dedicated to the review of empirical studies that control
for different market states when assessing financial performance.
2.1 An overview about SRI
Since the beginnings of the 90s the industry of SRI has been rising considerably
worldwide. The origins of SRI come from religious traditions and developed due to the
growing demand of products that were consistent with the consumers’ ethical values.
Due to a series of environmental disasters, social campaigns and posteriorly corporate
scandals, factors like Environmental, Social and Governance (ESG) turn out to be
important to investors when screening their investments (Renneboog et al., 2008a).
Commonly, SRI is seen as an investment in which decisions are based on ethical
and personal values instead of financial wealth (Derwall et al., 2011). But, according to
more recent studies, SR investors not only base their investment decision on the ethical
and personal factors, but also on risk-reward optimization to derive their utility function
from owning the securities. For instance, Nilsson (2009) identified three types of SRI
investors: those who base their investment decisions only on risk/return (“profit-
seeking” investor); those who base their investment decisions only on Social
Responsibility (“values-driven” investor); and finally, those who make their investment
decision simultaneously based on return and Social Responsibility.
According to USSIF (2018) one of the strategies that investors can follow to
engage in SRI, is to encourage firms to adopt Corporate Social Responsibility practices
through shareholders proposals. The other strategy, with more expression in the
market, is to incorporate ESG criteria to select a portfolio across a variety of asset
classes. An important segment of this strategy is to finance projects with the intent to
develop and help underserved communities.
17
SRI investors following ESG incorporation strategy can choose to invest in SRI
funds or construct themselves a portfolio consistent with their ideals. In order to fulfil
investors’ standards, different screen approaches can be implemented when building a
SR portfolio. The negative approach is the most basic and the oldest, which excludes
specific stocks or industries that do not rely on SRI ideologies (e.g. gambling, tobacco,
alcohol). The positive approach consists in selecting stocks that meet superior SRI
standards. Finally, is the Best-in-Class approach, where portfolios are constructed
selecting high rated SRI stocks from each industry. This last approach emerged with the
intent to overcome problems of sector biases and loss of diversification.
2.2 The performance of SRI versus the performance of conventional
strategies
Since the pioneering study of Moskowitz’s (1972), SRI has been widely studied
by empirical researchers in the last few years. Most studies on SRI seek to investigate
whether or not adding ESG criteria to the investment selection process has a positive
impact on portfolio returns. However, the conclusions of these studies are puzzling
because the expected performance of SR portfolios can be either lower, higher or equal
to the performance of conventional investments. Hence, different theoretical
arguments appeared in the literature in order to explain the impact of ESG criteria on
portfolio performance (see Hamilton and Statman, 1993; Bauer et al., 2005; Mollet and
Ziegler, 2014).
Following Markowitz’s (1952) portfolio theory, SRI portfolios have problems of
diversification and optimization, since they are constructed from a restricted universe
of investments, thus leading to lower performance. Traditionally, investors are assumed
to make investment decisions considering only the risk and return. This does not happen
with SRI investors (values-driven investor) since they shun controversial stocks that are
proved to present higher returns (Statman and Glushkov, 2009; Derwall et al., 2011;
Salaber, 2013). Derwall et al. (2011) discusses the shunned-stock hypothesis, which
states that relaxing the assumption of symmetrical information of the CAPM (Merton,
1987), the markets will segment due to different investors’ bases, which will affect stock
prices. Investors tend to invest more on certain stocks (SRI stocks) and neglect other
18
stocks (controversial stocks) that will be traded at a discount due to a low demand. On
the contrary, the increased demand for SRI stocks turns them overpriced generating
lower expected returns. Additionally, SRI investors incur in higher costs due to extra
informational needs, screening and monitoring processes (Bauer et al., 2005).
A more contemporary view, states that aligning all stakeholder’s interests,
including dimensions like social responsibility, creates more value for shareholders, and
improves financial performance (Waddock and Graves, 1997; Freeman et al., 2010). If
investors do not recognize it, the SRI stocks will be underpriced and, consequently, will
generate higher expected returns than conventional stocks. Furthermore, social screens
provide tools to select companies with higher management skills, and thus, SRI stock
portfolios will experience higher financial performance in the long run (Bollen, 2007).
The third argument is in line with an adaptation of the efficient capital market
theory made by Daniel and Titman (1999). It suggests that if it is possible to earn
abnormal returns using public information, this ability will disappear over time as soon
as this information is perceived by the market’s participants. Studies made by Bebchuk
et al. (2013) and Borgers et al. (2013) report positive abnormal returns for SRI in the
1990s. However, these abnormal returns became insignificant in subsequent years.
Their empirical results demonstrate that the mispricing of SRI gradually disappeared as
investors learned the advantages of ESG criteria. Therefore, SRI stocks should not be
mispriced and should not perform differently from their conventional counterparts.
2.3 Empirical evidence on the performance of SRI
Several researchers have been investigating the relationship between SR criteria
and financial performance. These studies have been developed in different lines (Cortez
et al., 2009). One body of the literature compares the financial performance of
companies that score good and bad in CSR. For instance, Orlitzy et al. (2003) and
Margolis and Walsh (2003) argue that financial performance is positively linked with
CSR. A second strand compares the performance of SR indices with conventional indices
and find that they do not perform differently (e.g. Sauer, 1997; Statman, 2006). The
following reviewed papers cover the third and the fourth strand of the literature that
19
analyses the performance of SR versus non-SR funds and the performance of portfolios
that scoring high versus low in ESG criteria, respectively.
Most researchers compare the performance of SR mutual funds with the
performance of conventional ones. There is little evidence that SR and conventional
funds perform differently. Studying worldwide SRI funds, Renneboog et al. (2008b) find
that SR funds underperform in the market; and specifically in France, Ireland, Sweden
and Japan, SR funds underperform their conventional counterparts. Additionally, in
Bauer et al. (2005), the results show that US international ethical funds underperform
their conventional counterparts for the period of 1990-1993, though US and UK
domestic SR funds outperform conventional funds for the period of 1994-1997 and
1998-2001, respectively.
However, in general, these types of studies show that the performance of SRI
funds are not statistically different from the performance of conventional funds. This
evidence is found in studies based on the US market (e.g. Hamilton et al., 1993; Reyes
and Grieb, 1998; Statman, 2000; Shank et al., 2005), the European market (e.g. Leite and
Cortez, 2014), on other, more specific markets (e.g. Bauer et al. 2007), as well as on
multi-country analyses (e.g. Bauer et al. 2005; Kreander et al. 2005; Cortez et al., 2009).
Nevertheless, this methodology presents some drawbacks as referenced by Auer
(2016). In fact, a labelled SRI fund does not always maintain a social responsibility status
as it is initially advertised. These changes are made by the manager due to other criteria
that do not rely on the level of Social Responsibility (Wimmer, 2013). A recent study
(Henke, 2016) revealed that one-third of the labelled SR bond funds, invest in bonds
with lower ESG ratings than conventional funds. Contrary to the “real” SR funds that
outperform the conventional ones, these “disguised” funds reveal no difference
financial performance when compared with conventional funds. This might explain the
fact that a significant number of studies, using this particular approach, have concluded
that SR funds’ performance is not different from the performance of conventional funds.
Moreover, Kempf and Osthoff (2007) point out that the financial performance of mutual
funds cannot be only attributed to the SRI returns, but it must also consider the fund
manager’s skills.
20
The other strand of literature that studies the performance of SRI uses synthetic
portfolios. Contrary to SR funds, these portfolios have the particularity to display the
effect of a particular ESG dimension on portfolio performance in isolation. This
methodology has two approaches regarding the ranking source. The portfolios can be
built using public lists or rankings of companies provided by rating agencies; or
alternatively, they can be built using rankings of firms constructed based on public ESG
scores of rating agencies. These types of studies have been presenting mixed results.
Even though there is evidence that SR portfolios deliver superior abnormal
performance, there is a lack of consensus in which public list, ESG dimension or screen
approach investors should rely on to build SR portfolios with good financial
performance.
Anderson and Smith (2006) find that constituent firms from the “America's Most
Admired Companies” list perform better than the market. Moreover, Statman et al.
(2008) and Angier and Statman (2010) find that the top companies underperform the
bottom companies of this list. Similar results were found by Preece and Filbeck (1999)
that analysed a portfolio composed by “100 Best Companies for Working Mothers”. This
portfolio, composed by firms on this list, outperforms the market, but underperforms
their matched sample. Although Filbeck et al. (2009) reached the same conclusion using
“The Best Corporate Citizens”, additionally, they also find that rebalancing the portfolio
each year, excluding the consecutive listed firms and including the new listed firms,
enables the portfolio to outperform the market and its matched sample.
Other studies demonstrate that a portfolio composed by “100 Best Companies
to Work for in America’’ outperform the market (Edmans, 2011) and its matched sample
portfolio (Filbeck and Preece, 2003; Filbeck et al., 2009). Using the same public list,
Carvalho and Areal (2016) find that some studies overestimated the performance of
these companies because they did not consider time-varying models. In their study, a
portfolio including all companies from the list, do not outperform the market while a
portfolio with the top half of companies do.
A study conducted by Filbeck et al. (2013) explores the four public lists
mentioned above. They state that using certain public rankings, (e.g. the “Best
Corporate Citizens” and the “Most Admired Companies to work for in America”), to form
21
SR portfolios yield higher returns. Companies that are listed in two or three rankings in
the same year produces incremental value. Moreover, only in the “Most Admired
Companies”, a company reselected in the subsequent year produces incremental value.
However, firms listed in “100 Best Companies to Work for in America” and “100 Best
Companies for Working Mothers” do not outperform their matched sample.
Following a different approach and focusing on the concept of “eco-efficiency”,
Derwall et al. (2005) measured the performance between two distinct SR stock
portfolios constructed based on corporate eco-efficiency scores over the period 1995-
2003. Their findings reveal that environmental criteria can substantially enhance the
performance of stock portfolios. The high-ranked portfolio outperforms the low-ranked
portfolio, and the positive difference between them cannot be explained by changes in
market sensitivity, investment style or industry bias even in the presence of transaction
costs. Likewise, Eccles et al. (2014) and Mollet et al. (2013) sustain that a portfolio
constructed with “High Sustainability” outperforms the market and “Innovators” firms,
outperform their matched non-SRI sample4.
Further studies account for multi SRI dimensions and strongly support that the
impact of each SRI dimension should be examined separately, since not all dimensions
deliver positive abnormal performance. Even so, there is no consensus on which
dimension or screen approach delivers a higher performance. That is the case of
Brammer (2006) who uses indicators from Ethical Investment Research Service and
three other studies that have used KLD’s SR indicators for different periods of time
(Kempf and Osthoff, 2007; Statman and Glushkov, 2009 and Galema et al., 2008).
Analysing employment, environment and community indicators for UK firms,
Brammer (2006) find that high scoring firms perform worse than non-scoring firms. Also,
positive returns are weakly associated with firms that scoring high in employment
4 Eccles et al. (2014) cautiously selected a portfolio from US “High Sustainability” firms from Asset4 and a matched sample of “Low Sustainability” firms and compared their performance. They constructed an equally-weighted index of all Sustainability Policies using Asset4 scores for 675 companies. For the top quartile firms, they investigated the historical origins of the policies conducting interviews, reading published reports and visiting firms’ websites. In the end, their “High Sustainability” portfolio was composed of 90 firms, which historical evidence had proved that these firms adopted a substantial number of these policies in the beginning of 1990s. Between 1993-2010, both portfolios outperform the markets but “High Sustainability” portfolio significantly outperforms the “Low Sustainability” ones. Mollet et al. (2013) studied the European “innovators” firms from Zurich Cantonal Bank(ZKB) and this firms also outperformed the market.
22
dimension, while negative returns are associated with environmental and community
responsible firms.
Kempf and Osthoff (2007) test multi SRI dimensions as well, but they also test
different screening approaches and cut-offs for the 1995-2003 period. Contrary to
negative screening, the positive and the Best-in-Class screening produces abnormal
returns following a Long-Short strategy for both equally and value-weighted portfolios.
They concluded that the highest abnormal returns can be achieved by adopting Best-in-
Class screening, when combining different ESG dimensions simultaneously and
restricting the portfolios to stocks with the highest scores. The evidence of abnormal
returns holds even after accounting for transaction costs.
The study of Statman and Glushkov (2009) analysed portfolios based on different
SR characteristics from 1992 to 2007. Their dataset distinguishes them from the study
conducted by Kempf and Osthoff (2007), because they excluded firms with no strength
or weakness indicators. They sustain that a Best-in-Class equally weighted high-ranked
portfolio can outperform a low-ranked portfolio when incorporating characteristics such
as community involvement, employee relations or overall performance. The value-
weighted portfolios also display positive abnormal returns in relation to employee
relations and overall performance. It is important to point out that the overall
outperformance appears to occur during the subperiod 1992-1999. Moreover, they
found evidence that the exclusion of shunned companies might generate disadvantages
that can offset the advantages of investing in companies with high ESG scores.
The third study using KLD data is Galema et al. (2008). Over the 1992-2006
period, they analysed SRI portfolios testing them in a General Methods of Moments
system, a system that allows the errors of equations to be correlated. In this context
only the equally-weighted community portfolio outperforms the market at a 10%
significant level. On the other hand, using value-weighted portfolios but at the same
significance level, only the employee relations dimension recorded positive abnormal
returns.
However, neither Mollet and Ziegler (2014) who analysed three portfolios
composed of European and US “sustainability leaders” firms from Morgan Stanley
23
Capital International (MSCI) and ZKB databases, nor Halbritter and Dorfleitner (2015)
who study SRI dimensions independently using three different sources of data, find that
high and low scoring portfolios perform differently even using a Best-in-Class approach.
It is important to notice that all the studies mentioned above agree that the
Carhart (1997) four-factor model is the most appropriate model to assess portfolio
performance. In fact, because they control for common investment styles, multifactor
models play an important role when assessing performance.
2.4 Investing in SRI: “the learning hypothesis”
Kempf and Osthoff (2007) try to understand whether the positive relationship
between ESG criteria and abnormal returns result from a temporary mispricing in the
market or not. However due to problems in the sample size, their outputs were not
significant.
From the 1990s until the beginning of the 2000s, we are able to find several
studies that give support to the benefit of investing in SRI portfolios. Although, more
recent studies report a decline in the positive abnormal returns of SRI portfolios (Derwall
et al., 2011; Bebchuk et al., 2013; Borgers et al., 2013 and Halbritter and Dorfleitner,
2015). These findings are consistent with “the learning hypothesis” presented by
Bebchuk et al. (2013), which states that investors gradually acknowledge the differences
between SR firms and non-SR firms. Once they fully incorporate that information, stocks
are correctly priced and the advantage of earning abnormal returns for SRI disappears.
Derwall et al. (2011) found that opportunities for different types of SR investors
coexist in the short-term, but for those that seek profit the opportunities fade in the
long-term. Analysing two distinct portfolios between 1992 to 2008 using KLD data, one
scoring high in employee relations and the other in controversial activities, only the
controversial portfolio maintains a stable positive and significant performance during all
subperiods. The portfolio rated high in employee relation showed, during the subperiod
1992-2006 and 1992-2008, a much lower and insignificant alpha.
As mentioned above, an increasing ESG awareness among investors, over time,
may result in a decreasing abnormal positive performance for SRI due to learning effects
24
in capital markets. With the purpose of testing this learning hypothesis, Borgers et al.
(2013) built a stakeholder-relation index and used three complementary methods: a
portfolio approach, an event study around earnings announcements and an analysis of
errors in analysts’ forecasts. All the methods confirmed that errors in expectations due
to lack of awareness existed during 1992-2004 but did not persisted during 2004-2009.
The positive abnormal performance and its statistical significance decreased in most of
the high-rated portfolios after 2004. Similarly, Bebchuk et al. (2013) reach the same
conclusions using Governance indices based on Investor Responsibility Research Center
data. They reported positive abnormal returns between 1990 and 1999 but these
positive abnormal performances became neutral in the subperiod of 2000-2008.
Nevertheless, both studies revealed that these indices are important tools for investors,
researchers and governance policymakers, since their relationship with firm value,
operating performance, and profit continued to persist overtime.
Studies mentioned above used different datasets and, therefore, the event of
“learning” was able to be captured, albeit with small variations, in different subperiods.
The comparison between different data sources of ESG scores carried out by Halbritter
and Dorfleitner (2015), does not find significant differences in the performance of high
and low-ranked US SRI portfolios. These results hold even when the Best-in-Class
strategy is used. However, when the dataset is divided into subperiods their results are
similar to those reported by the three previously mentioned studies. The positive
abnormal returns of equally weighted portfolios from KLD database prevailed during
1991-2001 but declined in the following years. The alphas based on the other data
sources revealed similar results, after the 2002-2006 period, they converge to zero.
2.5 SRI in Recession Periods
So, why does SRI demand continue to increase if they no longer generate more
positive returns? Although “values-driven” investors are willing to quit financial wealth
in order to have investments that reflect their convictions, “profit-seeking” investors do
not. Nofsinger and Varma (2014) suggest that the reason why SRI is in high demand is
possibly because investors want to minimize their downside risk and companies with
good Corporate Social Responsibility have characteristics that makes them less risky in
25
recession periods. In fact, the literature provides evidence that firms with strong
Corporate Social Responsibility activities reduce their litigation (Koh et al., 2014),
idiosyncratic (Godfrey et al., 2009; Ghoul et al., 2011; Bouslah et al., 2013) and stock-
price crash risks (Kim et al., 2014). Moreover, Bollen (2007) and Benson and Humphrey
(2007) found that SR funds’ flows are less sensitive to past negative returns than flows
of conventional funds. Thus, SR funds volatility is lower than conventional funds
volatility which might explain the fact that they perform better in bad times.
To understand this issue better, it is important to study the performance
accounting for different market conditions. The majority of literature that investigates
the performance of SR synthetic portfolios uses unconditional models to assess
performance, assuming that risk and return are constant over time. However, it is a well-
known fact that risk and return are not linear over time. In this regard, there are several
studies that evaluate performance, controlling for different market conditions.
Moskowitz (2000), Kosowski (2011) and Glode(2011), suggest that assessing
performance through unconditional models, may understate active managers’ abilities.
They assess financial performance during periods of recession and the results show
conventional equity mutual funds performing better in recession periods.
In respect to SRI, researchers have been reporting a positive relationship
between fund financial performance and recession periods. Although Nofsinger and
Varma (2014) found that, in general, conventional funds outperform the SRI funds, in
periods of crisis SRI funds outperform the conventional ones. They also conclude that
the positive alphas during the periods of crisis are associated with the positive screening
and ESG criteria. On the contrary, negative screening and criteria that focus on religion
or controversial activities lead funds to perform poorly during crisis periods. Similar
results are reported by Henke (2016) in relation to SRI bond funds. However, in this case,
the most successful strategy during the period of crisis is the exclusion of bond issuers
with low ESG scores from the bond mutual funds, instead of the inclusion of bonds with
higher ESG scores.
Additionally, Silva and Cortez (2016) analyse and compare the performance of
certified green funds, uncertified green funds and other SR funds. Although they find
that all US funds perform equally, green funds outperform the SRI funds in periods of
26
crises. Muñoz et al. (2014) evidences also shows that SR funds perform better in crisis
periods, outperforming their conventional peers.
To the best of my knowledge, Carvalho and Areal (2016) are the only ones
studying synthetic portfolio performance during different market states. They found
that companies from the “100 Best Companies to Work for in America” list, during a
market crisis, sustain their performance and, systematic risk and value, and the top
companies continue to outperform the market.
It is important to point out that when we are evaluating the performance of SRI
across different market conditions, the choice of the methodology used to define the
alternative market conditions may be critical, and the results of Areal et al. (2013)
support this. On the one hand, when defining market regimes based on market volatility,
the authors find that SRI mutual funds underperform during expansion periods and
slightly outperform during recession periods. On the other hand, when using NBER
business cycles, the financial performance of SRI does not change across different
market conditions.
27
3.Methodology and Dataset
This section describes the methodology implemented and the dataset used in
the empirical tests. Firstly, how ESG portfolios are constructed is explained. Following,
the unconditional models implemented to assess portfolio performance is described.
Then, market states are defined according to NBER business cycles and time-varying
models are presented. The final part of this section describes the ESG scores used to
construct the portfolios and the required data to assess the financial performance of
ESG portfolios.
3.1 ESG Portfolio Construction
One of the most common approaches in literature to analyse the effects of ESG
criteria on portfolio performance, is the construction of synthetic ESG portfolios. As
described by Halbritter and Dorfleitner (2015), this approach enables the aggregation of
a considerable amount of panel data in a single time-series dimension. This allows the
application of basic asset pricing models and it provides a straightforward trading
strategy for investors to exploit the relationship between ESG scores and the financial
performance. Also, as was previously stated, construction of synthetic portfolios based
on ESG scores to investigate the performance of SRI overcomes some limitations
associated with assessment of SRI performance based on SRI mutual funds. Therefore,
this section follows Kempf and Osthoff (2007), Statman and Glushkov (2009) and
Halbritter and Dorfleitner (2015).
Each month t from 2002 to 2017, two distinct equally-weighted portfolios for
each ESG dimension are constructed: High portfolios and Low portfolios. In month t-1
firms are ranked by their ESG scores. The portfolios are formed at the beginning of
month t and held until the end of month t. The 20% highest (lowest) scoring firms are
assigned accordingly to each dimension to the high (low) portfolio. Portfolios are
adjusted in a monthly basis.
The main focus is to analyse the impact of the ESG criteria on the financial
performance, and to accomplish that, a Long-Short strategy, which consists of holding
28
the High portfolio in a long position and the low portfolio in a short position, is followed.
Then, its performance is evaluated.
3.2 Performance Measurement
The simplest performance measure used in the literature is the Jensen’s (1968) alpha in
the context of the Capital Asset Pricing Model (CAPM), a one-factor model that only
accounts for the excess return of the market portfolio. However, this model has some
limitations related to the CAPM inefficiencies (e.g. Roll, 1977). To overcome such
limitations, Fama and French (1993) proposed the three-factor model, adding value and
size factors to the one-factor model. Later, Carhart (1997) made some improvements
adding the momentum factor of Jegadeesh and Titman’s (1993) suggesting that a four-
factor model displays more explanatory power than its predecessors. The Carhart (1997)
four-factor model is probably the most commonly used model in finance literature to
assess portfolio performance, including the performance of synthetic SRI portfolios (e.g.
Kempf and Osthoff, 2007; Derwall et al., 2005; Borgers et al., 2013). Therefore, the
performance of the ESG portfolios is initially assessed using the Carhart (1997) four-
factor model:
𝑅𝑖,𝑡 − 𝑅𝑓,𝑡 = 𝛼𝑖 + 𝛽1,𝑖(𝑅𝑚,𝑡 − 𝑅𝑓,𝑡) + 𝛽2,𝑖𝑆𝑀𝐵𝑡 + 𝛽3,𝑖𝐻𝑀𝐿𝑡 + 𝛽4,𝑖𝑀𝑂𝑀𝑡
+ 𝜀𝑖,𝑡
(1)
where 𝑅𝑖,𝑡 is the return on the portfolio 𝑖 in period 𝑡; 𝑅𝑓,𝑡 is the risk-free rate; 𝑅𝑚,𝑡 is
the return of the market portfolio; 𝑆𝑀𝐵𝑡 is the return difference between a small and a
large capitalisation portfolio in month t; 𝐻𝑀𝐿𝑡 the return difference between a
portfolio of high book-to-market stocks and a portfolio of low book-to-market stocks
and 𝑀𝑂𝑀𝑡 is the return difference between the portfolio of the past 12-month return
winners and losers. The 𝛽𝑠 measure the risk in respect to each factor and the Jensen’s
alpha, 𝛼𝑖 , measures the average abnormal return of an ESG portfolio in excess of the
return on the market portfolio.
Recently, Fama and French (2015) suggested an improved version of the Fama
and French (1993) three-factor model that adds profitability and investment factors: the
29
five-factor model. These authors claim that investors should choose the five-factor
model if they are interested in portfolios that tilt towards value, size, profitability and
investment premium. Since this is a quite recent model, it is of interest to test it to assess
the performance of ESG portfolios. The model is summarized by the following equation:
𝑅𝑖,𝑡 − 𝑅𝑓,𝑡 = 𝛼𝑖 + 𝛽1,𝑖(𝑅𝑚,𝑡 − 𝑅𝑓,𝑡) + 𝛽2,𝑖𝑆𝑀𝐵𝑡 + 𝛽3,𝑖𝐻𝑀𝐿𝑡 + 𝛽4,𝑖𝑅𝑀𝑊𝑡
+ 𝛽5,𝑖𝐶𝑀𝐴𝑡 + 𝜀𝑖,𝑡
(2)
where 𝑅𝑀𝑊𝑡 is the difference between the returns on diversified portfolios of stocks
with robust and weak profitability and 𝐶𝑀𝐴𝑡 is the difference between the returns on
diversified portfolios of the stocks of low and high investment firms.
3.3 Recession Periods
Both models presented above assume that portfolio performance and risk are
constant across different market conditions and may understate the ESG portfolio
performance. Some authors define market states based on the identification of periods
of high/low volatility in the stock market (e.g. Areal et al., 2013; Nofsinger and Varma,
2014). Other authors (Moskowitz, 2000; Kosowski, 2011; Areal et al., 2013 and Henke,
2016) distinguish between market states, using US National Bureau of Economic
Research (NBER) business cycles. NBER defines a recession when there is a significant
fall in the economic activity spread across the economy, that lasts for more than few
months5.
In this dissertation, the ESG portfolios performance is assessed across the
different NBER business cycles of recession and expansion. Figure 2 shows the monthly
evolution of the market portfolio from January 2002 to June 2017. The grey area in the
graphs identify the period of recession, and white areas correspond to the periods of
expansion according to NBER classification of business cycles. The only period of
recession identified in the graph begins in January 2008 and ends in June 2009,
5 According to the last announcement from the NBER’s Business Cycle Dating Committee from September 20 of 2010, the significant
fall in the economic activity is visible not only in the GDP, but also in real income, employment, industrial production and wholesale-retail sales.
30
corresponding to the latest global financial crisis and an accentuated decrease in the
stock market.
Figure 3, presented next, shows the monthly excess returns of the market
portfolio across the different NBER business cycles of recession and expansion. As
expected, the recession period represented in the graph is associated with high
volatility.
-18
-14
-10
-6
-2
2
6
10
14
Jan
-02
Oct
-02
Jul-
03
Ap
r-0
4
Jan
-05
Oct
-05
Jul-
06
Ap
r-0
7
Jan
-08
Oct
-08
Jul-
09
Ap
r-1
0
Jan
-11
Oct
-11
Jul-
12
Ap
r-1
3
Jan
-14
Oct
-14
Jul-
15
Ap
r-1
6
Jan
-17
EXC
ESS
RET
UR
NS
DATE
-40
-20
0
20
40
60
80
100
120
140
Jan
-02
Oct
-02
Jul-
03
Ap
r-0
4
Jan
-05
Oct
-05
Jul-
06
Ap
r-0
7
Jan
-08
Oct
-08
Jul-
09
Ap
r-1
0
Jan
-11
Oct
-11
Jul-
12
Ap
r-1
3
Jan
-14
Oct
-14
Jul-
15
Ap
r-1
6
Jan
-17
K.R
.F. M
arke
t P
ort
folio
DATE
Figure 3- Monthly excess returns of the US Market portfolio, available on the Professor Kenneth French webpage, over the period of January 2002 to September 2017.
Figure 2- Monthly evolution of the US Market Portfolio, available on the Professor Kenneth French webpage, over the period of January 2002 to September 2017.
31
Similarly, to Areal et al. (2013) and Carvalho and Areal (2016), to assess portfolio
performance across periods of expansion and recession a dummy variable, 𝐷𝑡, is added
to the four (equation 3) and five-factor (equation 4) models, based on the NBER business
cycle information:
𝑅𝑖,𝑡 − 𝑅𝑓,𝑡 = 𝛼𝑖 + 𝛼𝑟𝑒𝑐,𝑖𝐷𝑡 + 𝛽1,𝑖(𝑅𝑚,𝑡 − 𝑅𝑓,𝑡) + 𝛽1𝑟𝑒𝑐,𝑖(𝑅𝑚,𝑡 − 𝑅𝑓,𝑡)𝐷𝑡
+ 𝛽2,𝑖𝑆𝑀𝐵𝑡 + 𝛽2𝑟𝑒𝑐,𝑖𝑆𝑀𝐵𝑡𝐷𝑡 + 𝛽3,𝑖𝐻𝑀𝐿𝑡 + 𝛽3𝑟𝑒𝑐,𝑖𝐻𝑀𝐿𝑡𝐷𝑡
+ 𝛽4,𝑖𝑀𝑂𝑀𝑡 + 𝛽4𝑟𝑒𝑐,𝑖𝑀𝑂𝑀𝑡𝐷𝑡 + 𝜀𝑖,𝑡
(3)
For both models, the dummy variable assumes a value of 0 in periods of
expansion, and a value of 1 in periods of recession. It allows us to analyse differences
across market conditions, not only with respect to the alphas but also to the risk factors.
3.4. Dataset Description
ESG database, from Thomson Reuters DataStream is used to construct the ESG
portfolios. Its total universe comprises more than 3800 public firms worldwide with a
minimum of 4 years of history since 2002. Firms are scoring using more than 250
performance indicators calculated from more than 750 data points, which covers 4
different performance dimensions: Environmental, Social, Corporate Governance and
Economic. All firms are benchmarked against the rest of the firms in the database.
From ASSET4, the aggregated scores for the ENV, SOC and GOV dimensions are
extracted. The ENV score measures the impact of a firm’s activities in the ecosystems
and how well a company uses its management practices to avoid environmental risks
and use environmental opportunities to generate long-term shareholder value. The SOC
score measures the capacity of a company to generate trust and loyalty with
𝑅𝑖,𝑡 − 𝑅𝑓,𝑡 = 𝛼𝑖 + 𝛼𝑟𝑒𝑐,𝑖𝐷𝑡 + 𝛽1,𝑖(𝑅𝑚,𝑡 − 𝑅𝑓,𝑡) + 𝛽1𝑟𝑒𝑐,𝑖(𝑅𝑚,𝑡 − 𝑅𝑓,𝑡)𝐷𝑡
+ 𝛽2,𝑖𝑆𝑀𝐵𝑡 + 𝛽2𝑟𝑒𝑐,𝑖𝑆𝑀𝐵𝑡𝐷𝑡 + 𝛽3,𝑖𝐻𝑀𝐿𝑡 + 𝛽3𝑟𝑒𝑐,𝑖𝐻𝑀𝐿𝑡𝐷𝑡
+ 𝛽4,𝑖𝑅𝑀𝑊𝑡 + 𝛽4𝑟𝑒𝑐,𝑖𝑅𝑀𝑊𝑡𝐷𝑡 + 𝛽5,𝑖𝐶𝑀𝐴𝑡
+ 𝛽5𝑟𝑒𝑐,𝑖𝐶𝑀𝐴𝑡𝐷𝑡+𝜀𝑖,𝑡
(4)
32
stakeholders through its management practices to create shareholder value. The GOV
score evaluates how well a company manages its mechanism of incentives, which allies
its rights and responsibilities with those of the board members and CEOs, ensuring that
they act in the best interest to create long term shareholders’ value. The Economic
dimension has not been considered since its score reflects the company’s overall
financial health, which does not fit in the purpose of this dissertation.
Additionally, with the main purpose of analysing the performance of a SRI
portfolio constructed based on a score that reflects simultaneously the three
dimensions, an overall ESG score 6 is computed by taking the average of the
Environmental, Social and Governance score.
As the focus of this study is the financial performance of ESG portfolios, the ESG
data from ASSET4 was merged with the Thomson Reuters DataStream financial data. For
every company, the monthly total return index was obtained to calculate monthly
discrete returns. As a result, the monthly sample from January 2002 to September
20177, is comprised of 2355 US public firms, where scores and total return indexes were
available.
In the contexts of the Carhart (1997) four-factor and the Fama and French (2015)
five-factor models, all the factors were collected from the data library of Professor
Kenneth R. French’s website8, including the excess market return. Consequently, the
market portfolio is the value-weighted return of all CRSP firms incorporated in the US
and listed in the NYSE, AMEX and NASDAQ.
Table 1 reports the descriptive statistics for the returns of the ESG portfolios and
the market portfolio for a period of 188 months. All portfolios present mean returns
higher than the market portfolio, but the Low rated portfolios present higher values
than the High rated portfolios. As expected in the financial series of returns, the portfolio
returns present a negative skewness and excess kurtosis, thus none of the portfolios
present the normal distribution of the returns. The Jarque-Bera test, that confirms the
non-normality of the portfolio returns distribution was also performed.
6 ASSET4 provides an overall ESG score but it is based on the four dimensions which is inappropriate for this study. 7 The last time data were extracted (November 2017), the ASSET4 database was not updated and some months from 2017 didn’t have enough data to construct portfolios. To avoid biased estimations, the existing data from October 2017 forward, was excluded from the empirical procedures. 8 http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html
33
Table 1- Descriptive statistics of portfolios’ returns
Portfolio Max. Min. Mean Median Std. Dev.
Skewness Kurtosis JB Prob.
ENV
High 0.137 -0.170 0.009 0.014 0.045 -0.454 4.365 0.000
Low 0.135 -0.167 0.010 0.016 0.048 -0.416 3.680 0.011
SOC
High 0.123 -0.174 0.009 0.013 0.042 -0.545 4.761 0.000
Low 0.144 -0.180 0.012 0.018 0.049 -0.485 4.002 0.000
GOV
High 0.132 -0.177 0.009 0.013 0.046 -0.553 4.480 0.000
Low 0.162 -0.187 0.013 0.020 0.049 -0.468 4.245 0.000
ESG
High 0.123 -0.175 0.009 0.013 0.043 -0.503 4.614 0.000
Low 0.142 -0.182 0.013 0.018 0.048 -0.515 4.077 0.000
MKT 0.114 -0.172 0.007 0.012 0.041 -0.693 4.628 0.000
This table presents the descriptive statistics of portfolios returns constructed based on the positive screen approach. The high (low) portfolios are formed with the 20% highest (lowest) rated companies according to each ESG score. The maximum, minimum, mean, median, standard deviation, skewness, kurtosis, and the Jarque-Bera probability test of portfolios’ returns for each dimension are presented.
34
4. Results
In this section, the obtained results concerning the performance of SR portfolios
are reported and discussed 9 . First, the performance of High, Low and Long-Short
portfolios assessed using the four-factor and five-factor models are displayed. Next,
several robustness checks are implemented with respect to the weighted scheme,
screen approach, different cut-offs and the exclusion of financial firms. Finally, the
performance results of all portfolios assessed with models that allow performance and
risk to vary according to expansion versus recession periods is presented.
4.1 Performance Evaluation
Table 2 presents the performance estimates of the Carhart (1997) four-factor
model for High, Low and Long-Short equally weighted portfolios for each ESG dimension.
In terms of financial performance, the results show that Low and High rated portfolios
outperform the market with a significance level of 1% and 5%, except for the Low rated
portfolio of ENV dimension. However, Low rated portfolios display higher alphas than
High rated portfolios. Consequently, alphas of Long-Short portfolios are negative and
statistically significant at 1% and 5% level for GOV and ESG dimension. Moreover, High
and Low rated portfolios constructed in the basis of ENV and SOC dimensions do not
perform differently from each other. Overall, it is possible to conclude, that there is no
advantage to investing in a portfolio composed by High rated companies instead of a
Low rated one as the Long-Short strategy based on ESG scores, does not provide positive
abnormal returns.
The risk factors seem to explain well the excess returns of all equally weighted
portfolios, since its loadings are, in general, statistically significant at a 5% level with the
exception of the momentum factor for Low rated portfolios. Nevertheless, ENV and ESG
High rated portfolios are more exposed to market risk than Low rated portfolios while
the contrary is true for SOC and GOV dimensions. In every dimension, portfolios with
9 Tests are performed to analyse the presence of heteroscedasticity and autocorrelation. The White (1980) adjustment for the
presence of heteroscedasticity and the Newey West (1987) adjustment for the simultaneous presence of heteroscedasticity and autocorrelation are used.
35
high scoring firms are less exposed to size and book-to-market risk, which means that
High rated portfolios are less exposed to small capitalization stocks and to value firms
than Low rated portfolios.
Moreover, the momentum factor presents negative and statistically significant
coefficients at 1% and 5% significance level, indicating that portfolios are composed on
average by firms with poor past performance. These results are supported by the results
of Derwall et al. (2005), in which his portfolio composed by companies ranking high in
eco-efficiency also presents a negative momentum coefficient.
Table 2 - Portfolio performance estimates- Carhart (1997) four-factor model
Portfolio α Market SMB HML MOM R2
ENV
High 0.002*** 0.995*** 0.119*** 0.102*** -0.045*** 0.960
Low 0.002* 0.941*** 0.401*** 0.260*** -0.041 0.908
Long-Short -0.000 0.054 -0.282*** -0.158*** -0.004 0.186
SOC
High 0.002*** 0.962*** 0.081*** 0.052** -0.035** 0.961
Low 0.004*** 0.993*** 0.457*** 0.172*** 0.006 0.928
Long-Short -0.002 -0.030 -0.377*** -0.120*** -0.041 0.353
GOV
High 0.002** 0.998*** 0.187*** 0.089*** -0.058*** 0.957
Low 0.005*** 1.000*** 0.425*** 0.156*** -0.039* 0.941
Long-Short -0.003*** -0.002 -0.239*** -0.066 -0.019 0.186
ESG
High 0.002*** 0.967*** 0.100*** 0.075*** -0.050*** 0.963
Low 0.004*** 0.962*** 0.462*** 0.188*** -0.011 0.926
Long-Short -0.002** 0.004 -0.362*** -0.113** -0.038 0.308
This table presents the results of the Carhart (1997) four-factor model (Equation 1) from January 2002 to September 2017 on a monthly basis. The R2s, alphas, and factor loadings concerning market, size, value and momentum are reported. The high (low) portfolios are formed with the 20% highest (lowest) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. The portfolios are equally weighted. Standard errors were estimated using White (1980) or Newey-West (1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
Table 3 reports alphas and risk factors concerning the five-factor model for all
portfolios. Although the R2s have slightly increased, it can be said that the four and the
five-factor model have a very similar explanatory power. The main difference between
the results of the two models is the fact that alphas of High rated portfolios are not
36
statistically significant at a 5% level. A possible explanation for that might be the absence
of the MOM factor control, since High rated portfolios show significant exposure to this
factor. Comparatively to the four-factor model, the five-factor model, also displays a
negative and statistically significant Long-Short alpha for the SOC dimension, in addition
to GOV and ESG dimensions.
Table 3- Portfolio performance estimates- Fama and French (2015) five-factor model
Portfolio α Market SMB HML RMW CMA R2
ENV
High 0.001 1.042*** 0.128*** 0.080** 0.104** 0.061 0.960 Low 0.002* 0.935*** 0.384*** 0.233*** -0.088 -0.028 0.909
Long-Short -0.001 0.107*** -0.257*** -0.152*** 0.192*** 0.089 0.220 SOC
High 0.001* 1.014*** 0.090*** 0.019 0.134*** 0.103** 0.964 Low 0.004*** 0.988*** 0.474*** 0.128*** 0.020 -0.107 0.931
Long-Short -0.002** 0.026 -0.384*** -0.109** 0.114* 0.210*** 0.385 GOV
High 0.001 1.040*** 0.185*** 0.065 0.063 0.064 0.996 Low 0.005*** 1.012*** 0.433*** 0.138*** 0.007 -0.114* 0.943
Long-Short -0.004*** 0.028 -0.248*** -0.073* 0.056 0.178*** 0.217 ESG
High 0.001* 1.018*** 0.100*** 0.043 0.102*** 0.121*** 0.963 Low 0.004*** 0.967*** 0.483*** 0.149*** 0.032 -0.109 0.930
Long-Short -0.003*** 0.050* -0.383*** -0.106** 0.070 0.229*** 0.348
This table presents the results of the Fama and French (2015) five-factor model (Equation 2) from January 2002 to September 2017 on a monthly basis. The R2s, alphas, and factor loadings concerning market, size, value, investment and profitability are reported. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. The portfolios are equally weighted. Standard errors were estimated using White (1980) or Newey-West(1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
With respect to the risk factors, High rated portfolios show more exposure to the
market risk and less exposure to the size factor than Low rated portfolios, which
indicates that High rated portfolios are less exposed to small capitalization stocks than
Low rated portfolios. The book-to-market factor loadings have become not statistically
significant in most of the High rated portfolios. In fact, Fama and French (2015) explains
that “HML is redundant for describing average returns, because average HML return is
captured by exposures of HML to other factors” (p.12). Moreover, High rated portfolios
in the ENV, SOC and ESG dimension present positive and statistically significant
37
coefficients (at 1% and 5% level) for profitability factor. This indicates that these
portfolios are composed by profitable firms. The investment factor coefficients of SOC
and ESG Long-Short portfolios are positive at a 1% and 5% level, which means that these
portfolios are composed by firms that invest conservatively.
In a first approach, these results are in contrast with authors who argue that the
Long-Short strategy, based on ESG criteria, delivers positive returns (Derwall et al., 2005;
Kempf and Osthoff, 2007; Statman and Glushkov, 2009; Eccles et al., 2014). Moreover,
it is somehow in contrast with those authors who state that no significant return
difference exits between firms that scoring High and Low on ESG criteria (Brammer et
al., 2006; Halbritter and Dorfleitner, 2015). Although, Halbritter and Dornfleitner (2015)
conclude in their study that, between High and Low scoring ESG firms there is no
significant return difference, when analysing Asset4 equally-weighted portfolios, they
also find the GOV Long-Short portfolio to be delivering negative abnormal returns at a
5% significance level.
4.2 Robustnes checks: weightning scheme, screen approach, different cut-offs
and the exclusion of financial firms
Next some robustness checks are implemented to investigate if results still hold.
First, to investigate whether the results are dependent on the portfolio weighting
scheme, the financial performance of value weighted portfolios constructed based on
their Market Value is also measured. The alphas of the value weighted portfolios,
measured with the four-factor and five-factor model, are presented in table 4. The
performance estimates concerning High and Low rated portfolios can be found in
Appendix C and D. To enhance comparison, the alphas of equally weighted portfolios
are also displayed. The results are similar for both models. In general, the alphas of High
and Low rated value weighted portfolios are economically higher and statistically
significant at the 1% level. Also, differences between High and Low rated portfolios
increased, and, consequently, value-weighted Long-Short portfolios show more
negative and significant alphas at the 1% level, except for ENV dimension. The results of
value-weighted scheme strongly support that following a Long-Short strategy does not
deliver positive abnormal returns.
38
The second test follows some studies that state that a portfolio constructed with
a Best-in-Class screen approach results in superior performance for High rated portfolios
leading to a profitable Long-Short Strategy (Derwall et al., 2005; Kempf and Osthoff,
2007; Statman and Glushkov,2009). Thus, for every dimension, High, Low and Long-
Short portfolios are constructed based on this approach. In order to construct them,
each month t from 2002 to 2017, it was constructed two distinct equally- weighted
portfolios for each ESG dimension. In month t-1, firms are ranked by their scores. The
portfolios are formed in month t and held until the end of month t. The 20% highest
(lowest) scoring firms from each industry10 are assigned accordingly to each dimension
to the High (Low) portfolio. Once more, portfolios are rebalanced, and companies are
ranked in a monthly basis. Accordingly, in month t+1, the portfolios must be
reconstructed if a firm vanishes form the database.
Table 4 - Long-short portfolio performance estimates depending on the weighting scheme
Portfolio four-factor five-factor
Equally Value Equally Value
ENV 0.000 -0.002* -0.001 -0.003**
SOC -0.002 -0.004*** -0.002** -0.005***
GOV -0.003*** -0.004*** -0.004*** -0.005***
ESG -0.002** -0.006*** -0.003*** -0.006***
This table presents the Long-Short alphas of the Carhart (1997) four-factor model (Equation 1) and the Fama and French (2015) five-factor model (Equation 2) from January 2002 to September 2017 on a monthly basis. The portfolios are constructed on an equally and value weighted scheme based on their market value. The High (Low) portfolios are formed with the 20% High (Low) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West (1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
Table 5 reports the alphas estimates of the four and five-factor model for the
Positive and Best-in-Class screen approach. The performance estimates concerning High
and Low rated portfolios alphas are in Appendix F and G. The results for the Best-in-Class
screen approach did not change in terms of the value or the statistical significance for
10 Similarly, to Kempf and Osthoff (2007) industries were grouped in ten classes as defined in Professor Kenneth R. French website. See http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/Data_Library/det_10_ ind_port.html.
39
the both models. Significant negative abnormal returns continue to be delivered in both
models, when using a Best-in-Class screen approach for the GOV and ESG dimension and
also for the SOC dimension in the case of the five-factor model. In the same way,
although Halbritter and Dorfleintner (2015) ESG Long-Short portfolios present neutral
alphas, they also do not find greater benefits in following the Best-in-Class screen
approach. However, these results are counterintuitive with authors who argue in favour
of a Best-in-Class screen approach (Derwall et al., 2005; Kempf and Osthoff, 2007;
Statman and Glushkov, 2009).
Table 5 - Long-short portfolio performance estimates depending on the screen approach
Portfolio four-factor five-factor
Positive Best-in-Class Positive Best-in-Class
ENV 0.000 -0.001 -0.001 -0.002*
SOC -0.002 -0.002* -0.002** -0.002**
GOV -0.003*** -0.003*** -0.004*** -0.003***
ESG -0.002** -0.003*** -0.003*** -0.003***
This table presents the Long-Short alphas of the Carhart (1997) four-factor model (Equation 1) and the Fama and French (2015) five-factor model (Equation 2) from January 2002 to September 2017 on a monthly basis. The portfolios are equally weighted. The high (low) portfolios are formed with the 20% high (low) rated companies of each industry according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West (1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
In the third place, the portfolios were formed with three different cut-offs to find
whether the portfolios performance changes. The alphas of Long-Short portfolios are
presented in Table 6. The performance estimates concerning High and Low rated
portfolios’ alphas are in Appendix H and I. The results show no significant differences
when using different cut-offs. In general, most of the significant and insignificant Long-
Short alphas remained throughout the cut-offs. It is possible to observe, then, that the
return differences between High and Low rated portfolios is maintained in the 50%, 25%,
20% cut-offs. In the lowest cut-off the statistical significance of alphas drops but the
estimated alphas might be biased because the number of companies per portfolio
decreases. As seen in Halbritter and Dorfleitner (2015), there is no pattern concerning
40
value and sign, which contrasts with Kempf and Osthoff’s (2007) evidence that show
returns and their significance level increasing as the cut-offs decrease.
Table 6- Long-short portfolio performance estimates depending on the cut-off
Portfolio Cut-off
10% 20% 25% 50%
Panel A: 4-factor model
ENV 0.001 0.000 -0.001 0.000
SOC -0.002 -0.002 -0.001 -0.001
GOV -0.003** -0.003*** -0.003*** -0.002**
ESG -0.002 -0.002** -0.002** -0.001*
Panel B: 5- factor model
ENV 0.000 -0.001 -0.002* -0.001*
SOC -0.002 -0.002** -0.002** -0.001*
GOV -0.004** -0.004*** -0.003*** -0.002***
ESG -0.002* -0.003*** -0.003*** -0.002**
This table presents the Long-Short alphas of the Carhart (1997) four-factor model (Equation 1) and the Fama and French (2015) five-factor model (Equation 2) from January 2002 to September 2017 on a monthly basis. The portfolios are constructed on an equally weighted scheme according to 10%, 20%, 25% and 50% cut-off. The high (low) portfolios are formed with the 20% (10%, 25% and 50%) high (low) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West(1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
According to the “learning hypothesis”, SR portfolios should not present
abnormal returns after 2000. If this is true, then the ESG portfolios in the previous results
should not have presented abnormal returns. In order to find when the abnormal
returns occur, the dataset is divided into three subperiods.
Therefore, Table 7 presents the Long-Short alphas for the four and five-factor
model, estimated for three subperiods and the overall period. Performance estimates
of the High and Low rated portfolios’ alphas are in Appendix J and K. The results on panel
A and B show that in the first two subperiods the alphas of Long-Short portfolios display
neutral alphas except for the GOV dimension, which presents negative abnormal returns
at a 1% significance level in the first subperiod. In fact, these 2 first subperiods largely
correspond to the timeline where different authors find SR portfolios delivering neutral
alphas (Derwall et al., 2011; Bebchuk et al., 2013; Borgers et al., 2013 and Halbritter and
Dorfleitner, 2015). Halbritter and Dorfleitner (2015) also reported negative and
41
significant alpha in the first subperiod for the GOV dimension. Both models show that
the negative and significant Long-Short alphas of the overall period seem to have more
influence from the last subperiod.
Table 7- Long-short portfolio performance estimates for different subperiods
Long-Short Portfolio
Subperiod
2002-2006 2007-2011 2012-2017 Overall period
Panel A: 4-factor model ENV -0.001 0.001 -0.001 0.000
SOC -0.001 -0.001 -0.003 -0.002
GOV -0.004*** -0.002 -0.005** -0.003***
ESG -0.002 -0.001 -0.003* -0.002**
Panel B: 5-factor model ENV -0.002 0.002 -0.003 -0.001
SOC -0.001 -0.001 -0.004** -0.002**
GOV -0.004*** -0.002 -0.006*** -0.004***
ESG -0.002 -0.001 -0.005*** -0.003***
This table presents the Long-Short alphas of the Carhart (1997) four-factor model (Equation 1) and the Fama and French (2015) five-factor model (Equation 2) from January 2002 to September 2017 on a monthly basis. The sample was divided in three subperiods. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The portfolios are equally weighted. Standard errors were estimated using White (1980) or Newey-West (1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
Since financial firms are different from other sector firms in terms of their
valuation by the markets and their accounting rules (Mollet and Ziegler, 2014), in the
last analysis firms from the financial sector are excluded from the dataset. As in the
study of Mollet and Ziegler (2014), the results without financial firms strongly support
the previous ones. The Long-Short alphas of the four and five-factor model, for the full
sample, and the sample without financial firms, are put aside in Table 8. Performance
estimates concerning the High and Low rated portfolios alphas are in Appendix M and
N. The estimates of Long-Short alphas become more negative and significant in the case
of SOC and ESG dimensions in both models. Even in the ENV dimension the Long-Short
portfolio alpha becomes significantly negative at the 10% level when assessed with the
five-factor model. When the financial firms are excluded from the sample it becomes
more evident that the Low rated portfolios outperform the High rated portfolios, which
means that there is no advantage to following a Long-Short strategy.
42
In general, although it is possible to identify slight differences in statistical
significance of alphas, risk factors and R2, the two models present results that lead to
the same conclusions. Low rated portfolios outperform High rated portfolios making the
Long-Short strategy not profitable for the investors. The portfolios financial
performance maintains independently of the weighting scheme, screen approach or the
cut-off used and became stronger when excluding firms from the financial sector.
Table 8- Long-short performance estimates of portfolios with and without financial firms
Portfolio
four-factor five-factor
full sample subtracted sample
full sample subtracted sample
ENV 0.000 -0.001 -0.001 -0.002**
SOC -0.002 -0.002** -0.002** -0.003***
GOV -0.003*** -0.003*** -0.004*** -0.004***
ESG -0.002** -0.003*** -0.003*** -0.004***
This table presents the Long-Short alphas of the Carhart (1997) four-factor model (Equation 1) and the Fama and French (2015) five-factor model (Equation 2) from January 2002 to September 2017 on a monthly basis. All portfolios are equally weighted and the “subtracted sample” does not include firms from the financial sector. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The portfolios are equally weighted. Standard errors were estimated using White (1980) or Newey-West (1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
4.3 Performance in expansion versus recession periods
Table 9 presents the performance estimates of the four-factor model where EX,
represents the alphas and risk factors for expansion periods, and Dt represents the
corresponding dummies, that aim to assess if portfolios perform differently in recession
periods (in comparison to expansion periods)11.
Focusing on dummies, only the GOV Long-Short portfolio presents a positive and
statistically significant alpha at the 1% level for the recession period. Although alphas of
the recession period in the High and Low rated portfolios from the GOV dimension
suggest that their performances do not change significantly in recession periods, the
11 From the Appendix O to V, robustness checks are also implemented with respect to the weighted scheme, screen approach, different cut-offs and the exclusion of financial firms for the models with dummies.
43
GOV Long-Short portfolio suggests that in periods of recession, the High rated portfolio
performs significantly better than the Low rated portfolio. Most of the risk factors for
the recession period are not statistically significant. The main exception is for the
momentum factor in where the loadings concerning Low rated portfolios become
negative or more significantly negative in recession periods. This means that in recession
periods, Low rated portfolios are also composed by firms with poor past performance.
Consequently, in general, the momentum factor for the recession periods of the Long-
Short portfolios, show that differences between High and Low rated portfolios become
significantly positive in recession periods.
Table 10 presents the alphas and risk factors coefficients, for periods of
expansion (EX) and the corresponding dummies of recession (Dt) for the five-factor
model. Contrary to the four-factor model, the alphas for the recession period show that
High rated portfolios perform significantly better in recession periods, since their
coefficients are positive and statistically significant, but none of the Long-Short
portfolios presents significant alphas for the recession period. This means that, on the
context of the five-factor model, there is no advantage for an investor to follow a Long-
Short strategy based on ESG criteria even if the focus is to “survive” the adversities
during recession periods.
In relation to the risk factors, most of the risk factor for the recession periods are
not statistically significant, meaning that portfolios present approximately the same
exposure in recession periods as in expansion periods.
44
Table 9- Portfolio performance estimates - Carhart (1997) four-factor model with dummies
Portfolio Α Market SMB HML MOM
R2
EX Dt EX Dt EX Dt EX Dt EX Dt
ENV
High 0.001 0.005* 1.014*** -0.005 0.100*** 0.027 0.145*** -0.128** -0.032 -0.018 0.962 Low 0.001 -0.000 0.986*** -0.172** 0.416*** -0.413** 0.273*** -0.016 -0.014 -0.126** 0.916
Long-Short -0.000 0.005 0.028 0.167* -0.316*** 0.440** -0.128** -0.112 -0.018 0.108* 0.246 SOC
High 0.002** 0.004* 0.969*** 0.034 0.063** 0.083 0.082*** -0.097 -0.025 0.001 0.963 Low 0.003*** -0.001 1.017*** -0.082 0.449*** -0.118 0.216*** -0.175* 0.047 -0.137*** 0.932
Long-Short -0.002 0.006 -0.048 0.116 -0.385*** 0.201 -0.134*** 0.078 -0.072** 0.138** 0.386 GOV
High 0.001* 0.002 1.003*** 0.036 0.173*** 0.040 0.156*** -0.239*** -0.056** -0.007 0.960 Low 0.005*** -0.005 1.006*** -0.051 0.425*** -0.120 0.188*** -0.165* 0.007 -0.154*** 0.945
Long-Short -0.004*** 0.007*** -0.004 0.087 -0.252*** 0.160 -0.031 -0.074 -0.063 0.147*** 0.247 ESG
High 0.002** 0.005* 0.971*** 0.040 0.089*** 0.039 0.100*** -0.073 -0.049** 0.022 0.964 Low 0.004*** -0.001 0.990*** -0.104 0.467*** -0.293* 0.212*** -0.099 0.037 -0.161*** 0.932
Long-Short -0.002** 0.006 -0.019 0.144* -0.378*** 0.332* -0.112** 0.026 -0.086*** 0.184*** 0.368
This table presents the results of the Carhart (1997) four-factor model with dummies (Equation 3) from January 2002 to September 2017 on a monthly basis. The R2s, alphas, and factor loadings concerning market, size, value and momentum are reported for periods of expansion (RE) and recession (Dt). Periods of recession and expansion were defined using NBER business cycles. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The portfolios are equally weighted. Standard errors estimated using White (1980) or Newey-West (1987). ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
45
Table 10- Portfolio performance estimates - Fama and French (2015) five-factor model with dummies
Portfolio α Market SMB HML RMW CMA
R2 EX Dt EX Dt EX Dt EX Dt EX Dt EX Dt
ENV
High 0.000 0.009*** 1.052*** -0.006 0.114*** 0.005 0.111*** -0.102 0.108*** -0.225 0.065 0.046 0,963
Low 0.00 0.007 0.973*** -0.019 0.410*** -0.428** 0.262*** 0.039 -0.078 -0.041 -0.098 0.370 0,914
Long-Short -0.001 0.002 0.079* 0.012 -0.297*** 0.433** -0.151** -0.141 0.186** -0.183 0.163* -0.324 0,271
SOC
High 0.001 0.007** 1.010*** 0.008 0.077*** 0.048 0.025 -0.055 0.139*** -0.203 0.136*** -0.092 0,966
Low 0.003*** -0.002 1.005*** 0.021 0.481*** -0.136 0.200*** -0.136 0.009 0.291 -0.168** -0.023 0,934
Long-Short -0.003** 0.008* 0.005 -0.013 -0.404*** 0.184 -0.174*** 0.082 0.130** -0.494* 0.303*** -0.069 0,433
GOV
High 0.001 0.007** 1.039*** 0.009 0.175*** 0.028 0.118** -0.192*** 0.072 -0.266 0.076 -0.087 0,959
Low 0.005*** -0.000 1.009*** 0.045 0.444*** -0.115 0.173*** -0.104 0.009 0.003 -0.132* 0.044 0,944
Long-Short -0.005*** 0.007 0.030 -0.036 -0.269*** 0.143 -0.056 -0.088 0.063 -0.269 0.208*** -0.132 0,249
ESG
High 0.001 0.007** 1.013*** 0.001 0.089*** 0.027 0.040 -0.015 0.108*** -0.203 0.163*** -0.164 0,966
Low 0.004*** 0.003 0.987*** 0.048 0.506*** -0.337** 0.196*** -0.065 0.030 0.105 -0.176** 0.320 0,933
Long-Short -0.003*** 0.005 0.026 -0.047 -0.417*** 0.364** -0.155*** 0.050 0.078 -0.307 0.339*** -0.484 0,410
This table presents the results of the Fama and French (2015) five-factor model with dummies (Equation 4) from January 2002 to September 2017 on a monthly basis. The R2s, alphas, and factor loadings concerning market, size, value investment and profitability are reported for periods of expansion (RE) and recession (Dt). Periods of recession and expansion were defined using NBER business cycles. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The portfolios are equally weighted. Standard errors were estimated using White (1980) or Newey-West (1987). ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
46
5. Conclusion
This dissertation studies how screening a portfolio using ESG criteria can impact
portfolio financial performance. To do so, for each ESG dimension, it is constructed and
assessed the performance of two distinct portfolios, and the Long-Short strategy
between January 2002 and September 2017 is tested. To the best of my knowledge,
there seems to not exist many empirical evidences evaluating the financial performance
of SR synthetic portfolios, that had study each ESG criteria while considering different
market states. To overcome this issue, a dummy based on NBER business cycles to
account for different market conditions, it was added to the Carhart (1997) and Fama
and French (2015) models.
The literature gives evidence that portfolios constructed based on SR criteria can
deliver positive abnormal returns. However, there is no consensus on which approach
concerning the construction of the SR portfolio is most beneficial. Moreover, according
to the “learning hypothesis” (Bebchuk et al., 2013), the advantage of getting positive
abnormal returns from SR portfolios disappeared in the beginning of the 2000’s. Some
authors state that the reason why investors continue to choose SRI is because SR
companies, due to its SR characteristics, are able to perform better in worst times like
economic recessions.
The results suggest that if investors pursue a Long-Short strategy based on the
GOV or ESG dimension (and SOC in the case of the five factor-model), they are going to
obtain negative abnormal returns. The results are robust using the value weighted
scheme, the Best-in-Class screen approach, different cut-offs and when excluding the
financial firms. The division of the dataset into three subperiods reveals that between
2002 and 2011, the Long-Short strategy does not deliver significant abnormal returns.
This is in accordance to the literature (Derwall et al., 2011; Bebchuk et al., 2013; Borgers
et al., 2013 and Halbritter and Dorfleitner, 2015). However, between 2012 and 2017,
empirical results suggest that the tendency changes and it seems that the negative
abnormal returns obtained in the full dataset derives from the last subperiod.
When a dummy is applied to enable a different performance in times of
expansion and recession, the for the four-factor model gives evidence that a GOV Long-
Short portfolio delivers positive abnormal returns during recession periods. This
47
suggests that GOV SR firms, due to their characteristics and good GOV, perform better
in times of recession. However, this result is not consistent when using the five-factor
model and other robustness tests. Moreover, when financial firms are excluded from
the dataset, none of the portfolios present significant changes in alphas in recession
period.
Overall, the results suggest that investors pursuing a Long-Short strategy based
on ESG criteria can expect negative abnormal returns if they tilt to the GOV, ESG and
perhaps SOC dimension independently of the market state. These results are in line with
Carvalho and Areal (2016) studies, in the sense that ESG portfolios’ performances
maintains, independently of the market state. However, contrary to what a range of the
literature says (Nofsinger and Varma, 2014; Muñoz et al., 2014; Henke, 2016; Silva and
Cortez, 2016), there is no advantage to investing in SR portfolios if the goal is to survive
in times of recession. If their performance is equal in times of recession and expansion,
they will continue to have a neutral or negative performance (depending on dimension
they are based on).
Nevertheless, this study has some limitations that would be of interest for future
research. Firstly, this dissertation fails to study other markets. Secondly, since both High
and Low rated portfolios present positive abnormal returns, it would be worth the time
to construct a Long-Short strategy where investors would trade Long ESG portfolios with
high scoring firms and Short portfolios of conventional stocks. Finally, Areal et al. (2013)
suggests that the way market states are define on the methodology can influence
results, and, therefore, researchers should use other methodologies that also account
for risk and returns that vary over time. Besides these limitations the present study
provides evidence that financial performance of SR portfolios seems to be changing.
Since, between 2012 and 2017, portfolios present negative financial performance,
researchers could analyse different sources of ESG score data as Halbritter and
Dorfleitner (2015) did, to understand if these results are transversal to other data
sources and if the financial performance tendencies of SR portfolio are in fact changing.
48
References
- Anderson, J., & Smith, G. (2006). A Great Company Can Be a Great Investment. Financial
Analysts Journal, 62(4), 86-93.
- Angier, D., & Statman, M. (2010). Stocks of Admired and Spurned Companies. The Journal
of Portfolio Management, 36(3), 71-77
- Areal, N., Cortez, M., & Silva, F. (2013). The conditional performance of US mutual funds
over different market regimes: do different types of ethical screens matter? Financial
Market and Portfolio Management, 27(4), 397-429.
- Auer, B. R. (2016). Do Socially Responsible Investment Policies Add or Destroy European
Stock Portfolio Value? Journal of Business Ethics, 135(2), 381-397.
- Bauer, R., Koeddijk, K., & Otten, R. (2005). International evidence on ethical mutual fund
performance and investment style. Journal of Baking and Finance, 29(7), 1751-1767.
- Bauer, R., Derwall, J., & Otten, R. (2007). The Ethical Mutual Fund Performance Debate: New
Evidence from Canada. Journal of Business Ethics, 70(2), 111-124.
- Bebchuk, L., Cohen, A., & Wang, C. (2013). Learning and the disappearing association
between governance and returns. Journal of Financial Economics, 108(2), 323-348.
- Benson, K., & Humphrey, J. (2007). Socially responsible investment funds: Investor reaction
to current and past returns. Journal of Banking and Finance, 32, 1850–1859.
- Borgers, A., Derwall, J., & Koedijk, K. (2013). Stakeholders relations and stock returns: On
errors in investors’ expectations and learning. Journal of Empirical Finance, 22, 159-175.
- Bollen, N. (2007). Mutual Fund Attributes and Investor Behavior. Journal of Financial and
Quantitative Analysis, 42(3), 683-708.
- Bouslah, K., Kryzanowski L., & M’Zali B. (2013). The impact of the dimensions of social
performance on firm risk. Journal of Banking and Finance, 37, 1258–1273.
- Brammer, S., Brooks, C., & Pavelin, S. (2006). Corporate Social Performance and Stock
Returns: UK Evidence from Disaggregate Measures. Financial Management, 35, 97-116.
- Carhart, M. (1997). On the persistence in mutual fund performance. Journal of Finance, 52,
57-82.
- Carvalho, A., & Areal, N. (2016). Great places to work: Resilience in times of crisis. Human
Resource Management, 55(3), 479-498.
- Cortez, M.C., Silva, F., & Areal, N. (2009). The performance of European socially responsible
funds. Journal of Business Ethics, 87, 573–588.
49
- Daniel, K., & Titman, S. (1999). Market efficiency in an irrational world. Financial Analysts
Journal, 55(6), 28–40.
- Derwall, J., Guenster, N., Bauer, R., & Koedijk, K. (2005). The eco-efficiency premium puzzle.
Financial Analysts Journal, 61(2), 51-63.
- Derwall, J., Koedijk, K., & Horst, J. (2011). A tale of values-driven and profit-seeking social
investors. Journal of Banking and Finance, 35, 2137-2147.
- Eccles, R. G., Ioannou, I., & Serafeim, G. (2014). The impact of corporate sustainability on
organizational processes and performance. Management Science, 60(11), 2835-2857.
- Edmans, A. (2011). Does the stock market fully value intangibles? Employee satisfaction and
equity prices. Journal of Finance Economics, 101, 621-640.
- Fama, E., & French, K. (1993). Common risk factors in the returns on bonds and stocks.
Journal of Financial Economics, 33, 3-53.
- Fama, E., & French, K. (2015). A five-factor asset pricing model. Journal of Financial
Economics, 116, 1-22.
- Filbeck, G., & Preece, D. (2003). Fortune’s Best 100 Companies to Work for in America: Do
They Work for Shareholder? Journal of Business Finance and Acounting, 30 (5), 771-797.
- Filbeck, G., Gorman, R., & Zhao, X. (2009). The “Best Corporate Citizens”: Are They Good for
Their Shareholders? The Financial Review, 44, 239-262.
- Filbeck, G., Gorman, R., & Zhao, X. (2013). Are the best of the best better than the rest? The
effect of multiple rankings on company value. Review of Quantitative Finance and
Accounting, 41(4), 695-722.
- Freeman, R., Harrison, J., Wicks, A., Parmar, B., & Colle,S. (2010). Stakeholder Theory- The
State of the Art. New York: Cambridge University Press.
- Galema, R., Plantinga, A., & Scholtens, B. (2008). The stocks at stake: Return and risk in
socially responsible investment. Journal of Banking and Finance, 32(12), 2646-2654.
- Ghoul, S., Guedhami, O., Kwok, C., & Mishra, D. (2011). Does corporate social responsibility
affect the cost of capital? Journal of Banking and Finance, 35, 2388–2406.
- Godfrey, P., Merrill, C., & Hansen, J. (2009) The Relationship Between Corporate Social
Responsibility and Shareholder Value: An Empirical Test of the Risk Management
Hypothesis. Strategic Management Journal, 30, 425–445.
- Glode, V. (2011). Why mutual funds ‘underperform’? Journal of Financial Economics, 99,
546-559.
- Halbritter, G., & Dorfleitner, G. (2015). The wages of social responsibility—where are they?
A critical review of ESG investing. Review of Financial Economics, 26, 25-35.
50
- Hamilton, S., Jo, H., & Statman, M. (1993). Doing Well While Doing Good? The Investment
Performance of Socially Responsible Mutual Funds. Financial Analysts Journal, 49(6), 62-66.
- Henke, H. (2016). The effect of social screening on bond mutual fund performance. Journal
of Banking and Finance, 67, 69-84.
- Jegadeesh, N., & Titman, S. (1993), Returns to buying winners and selling losers: Implications
for stock market efficiency. Journal of Finance, 48, 65-91.
- Jensen, M. (1968). The performance of mutual funds in the period 1945-1965. Journal of
Finance, 23(2), 389-416.
- Kempf, A., & Osthoff, P. (2007). The effect of socially responsible investing on portfolio
performance. European Financial Management, 13(5), 908–922.
- Kim, Y., Li, H., & Li, S. (2014). Corporate social responsibility and stock price crash risk. Journal of
Banking and Finance, 43, 1–13.
- Koh, P., Qian, C., & Wang, H. (2014). Firm Litigation Risk and The Insurance Value of
Corporate Social Performance. Strategic Management Journal, 35, 1464–1482.
- Kosowski, R. (2011). Do mutual funds perform when it matters most to investors? US mutual
fund performance and risk in recessions and expansions. Quarterly Journal of Finance, 3,
607-664.
- Kreander, N., Gray, R., Power, D., & Sinclair, C. (2005). Evaluating the performance of ethical
and non-ethical funds: A matched pair analysis. Journal of Business Finance, 32(7) & (8),
1465-1493.
- Leite, P., & Cortez, M. (2014). Style and performance of international socially responsible
funds in Europe. Research in International Business and Finance, 30, 248-267.
- Margolis, J. & Walsh, J. (2003). Misery Loves Companies: Rethinking Social Initiatives by
Business. Administrative Science Quarterly, 48, 268–305.
- Markowitz, H. (1952). Portfolio Selection. The Journal of Finance, 7(1), 77-91.
- Merton, R. (1987). A Simple Model of Capital Market Equilibrium with Incomplete
Information. The Journal of Finance, 42(3), 483-510
- Mollet, J., & Ziegler, A. (2014). Socially responsible investing and stock performance: New
empirical evidence for the US and European stock markets. Review of Financial Economics,
23(4), 208-216.
- Mollet J., Arx, U., & Ilic´ D. (2013) Strategic sustainability and financial performance:
exploring abnormal returns. Journal of Business Economics, 83, 577-604.
- Moskowitz, M. (1972). Choosing Socially Responsible Stocks. Business and Society Review,
1, 71–75.
51
- Moskowitz, T. (2000). Discussion: mutual fund performance: an empirical decomposition
into stock-picking talent, style, transaction costs, and expenses. The Journal of Finance, 55,
1655-1703.
- Muñoz, F., Vargas, M., & Marco, I. (2014). Environmental Mutual Funds: Financial
Performance and Managerial Abilities. Journal of Business Ethics, 124, 551-569.
- National Bureau of Economic Research (NBER), 2012. Business Cycles Data.
<http://www.nber.org/cycles.html>.
- Newey, W., & West, k. (1987). A simple, positive semi-definite, heteroskedasticity and
autocorrelation consistent covariance matrix. Econometrica, 55, 703–708.
- Nilsson, J. (2009). Segmenting socially responsible mutual fund investors: the influence of
financial return and social responsibility. International Journal of Bank Marketing, 27(1), 5-
31.
- Nofsinger, J., & Varma, A. (2014). Socially responsible funds and market crises. Journal of
Banking and Finance, 48, 180-193.
- Orlitzky, M., Schmidt, F. & Rynes, S. (2003). Corporate Social and Financial Performance: A
Meta-Analysis, Organization Studies 24(3), 403–441.
- Preece, D., & Filbeck, G. (1999). Family Friendly Firms: Does It Pay to Carer? Financial
Services Review, 8, 47-60.
- Renneboog, L., Horst, J., & Zhang, C. (2008a). Socially Responsible investments: Institutional
aspects, performance and investor behaviour. Journal of Banking and Finance, 32, 1723-
1742.
- Renneboog, L., Horst, J., & Zhang, C. (2008b). The price of ethics and stakeholder
governance: the performance of socially responsible mutual funds. Journal of Corporate
Finance, 14, 302-322.
- Reyes, M., & Grieb, T. (1998). The External Performance of Socially-Responsible Mutual
Funds. American Business Review, 16(1), 1-7.
- Roll, R. (1977). A critique of the asset pricing theory's tests. Part I: on past and potential
testability of the theory. Journal of Financial Economics, 4(2), 129-176.
- Salaber, J. (2013). Religion and returns in Europe. European Journal of Political Economy, 32,
149-160.
- Sauer, D. (1997). The Impact of Social-Responsibility Screens on Investment Performance:
Evidence from the Domini 400 Social Index and Domini Equity Mutual Fund. Review of
Financial Economics 6(2), 137–149.
52
- Shank, T., Manullang, D., & Hill, R. (2005). “Doing well while doing good” revisited: a study
of socially responsible firms’ short-term versus long-term performance. Managerial Finance,
30(8), 33-46.
- Silva, F., & Cortez, M. (2016). The performance of US and European green funds in different
market conditions. Journal of Cleaner Production, 135, 558-566.
- Statman, M. (2000). Socially Responsible Mutual Funds. Financial Analysts Journal, 56(3),
30–39.
- Statman, M. (2006). Socially Responsible Indexes: Composition, Performance and Tracking
Error. Journal of Portfolio Management 32(3), 100–109.
- Statman, M., Fisher, K., & Anginer, D. (2008). Affect in a Behavioral Asset-Pricing Model.
Financial Analysts Journal, 64(2), 20-29
- Statman, M., & Glushkov, D. (2009). The Wages of Social Responsibility. Financial Analysts
Journal, 65(4), 33-46.
- Waddock, S., & Graves, S. (1997). The Corporate Social Performance-Financial Performance
Link. Strategic Management Journal, 18(4), 303-319.
- White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct
test for heteroskedasticity. Econometrica 48(4), 817–838.
- Wimmer, M. (2013). ESG-persistence in socially responsible mutual funds. Journal of
Management and Sustainability, 3(1), 9–15.
53
Appendix
Appendix A - Definition of Market States according to NBER Business cycles
Start End Market State Dummy
31-Jan-02 31-Dec-07 Expansion 0
31-Jan-08 30-Jun-09 Recession 1
31-Jul-09 31-Set-17 Expansion 0
In this table are resumed the periods of expansion and recession between January 2002 to September 2017 according to the Business Cycles of NBER. For periods of expansion the Dummy of time-varying models assumes a value of 0 in expansion periods and 1 in recession periods.
54
Appendix B - Descriptive statistics of value weighted portfolios returns
Portfolio Max. Min. Mean Median Std. Dev. Skewness Kurtosis JB Prob.
ENV
High 0.105 -0.124 0.010 0.014 0.038 -0.368 3.677 0.020
Low 0.134 -0.149 0.013 0.017 0.043 -0.275 3.965 0.008
SOC
High 0.102 -0.140 0.009 0.014 0.037 -0.626 4.391 0.000
Low 0.137 -0.163 0.015 0.016 0.045 -0.387 3.823 0.007
GOV
High 0.106 -0.137 0.010 0.013 0.038 -0.508 4.101 0.000
Low 0.143 -0.170 0.015 0.019 0.045 -0.369 4.301 0.000
ESG
High 0.101 -0.128 0.009 0.013 0.037 -0.472 4.030 0.000
Low 0.136 -0.160 0.016 0.020 0.044 -0.358 3.902 0.006
MKT 0.114 -0.172 0.007 0.012 0.041 -0.693 4.628 0.000
This table presents the descriptive statistics of value-weighted portfolios returns constructed based on the positive screen approach. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The maximum, minimum, mean, median, standard deviation, skewness, kurtosis, and the Jarque-Bera probability test of portfolios’ returns for each dimension.
Appendix C – Portfolio performance estimates of the value-weighted scheme– Carhart (1997) four-factor model
Portfolio α Market SMB HML MOM R2
ENV High 0.004*** 0.919*** -0.212*** 0.003 -0.015 0.947
Low 0.006*** 0.937*** 0.172*** 0.033 -0.026 0.901
Long-Short -0.002* -0.018 -0.384*** -0.030 0.011 0.245
SOC High 0.003*** 0.908*** -0.233*** 0.001 -0.004 0.945
Low 0.007*** 0.968*** 0.314*** -0.011 0.006 0.904
Long-Short -0.004*** -0.059 -0.547*** 0.013 -0.009 0.393
GOV High 0.003*** 0.928*** -0.161*** -0.049 0.009 0.936
Low 0.008*** 0.986*** 0.166*** -0.072 -0.025 0.896
Long-Short -0.004*** -0.058 -0.327*** 0.023 0.033 0.202
ESG High 0.003*** 0.895*** -0.231*** -0.016 0.002 0.939
Low 0.009*** 0.931*** 0.307*** 0.031 0.021 0.880
Long-Short -0.006*** -0.036 -0.539*** -0.047 -0.019 0.323
This table presents the estimates of the Carhart (1997) four-factor model (Equation 1) from January 2002 to September 2017 on a monthly basis. The R2s, alphas, and factor loadings concerning market, size, value and momentum are reported. The portfolios are constructed on a value weighted scheme. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West (1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
-
55
Appendix D – Portfolio performance estimates of the value-weighted scheme – Fama and French (2015) five-factor model
Portfolio α Market SMB HML RMW CMA R2
ENV High 0.003*** 0.944*** -0.225*** -0.007 0.040 0.177*** 0.951
Low 0.006*** 0.932*** 0.159*** 0.024 -0.062 -0.001 0.901
Long-Short -0.003** 0.012 -0.384*** -0.031 0.103 0.178** 0.273
SOC High 0.003*** 0.933*** -0.241*** -0.009 0.061* 0.165*** 0.950
Low 0.007*** 0.970*** 0.343*** -0.021 0.057 -0.173** 0.908
Long-Short -0.005*** -0.038 -0.584*** 0.012 0.004 0.338*** 0.452
GOV High 0.003*** 0.943*** -0.180*** -0.098** 0.034 0.264*** 0.939
Low 0.008*** 0.994*** 0.181*** -0.048 0.017 -0.159** 0.899
Long-Short -0.005*** -0.052 -0.361*** -0.050 0.017 0.423*** 0.290
ESG High 0.003*** 0.918*** -0.243*** -0.042 0.058 0.217*** 0.947
Low 0.009*** 0.932*** 0.353*** 0.017 0.084 -0.191** 0.887
Long-Short -0.006*** -0.014 -0.595*** -0.058 -0.026 0.408*** 0.400
This table presents the estimates of the Fama and French (2015) five-factor model (Equation 2) from January 2002 to September 2017 on a monthly basis. The R2s, alphas, and factor loadings concerning market, size, value, investment and profitability are reported. The portfolios are constructed on a value weighted scheme. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West (1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
56
Appendix E - Descriptive statistics of Best-in-Class portfolios returns
Portfolio Max. Min. Mean Median Std.Dev. Skewness Kurtosis JB Prob.
ENV
High 0.144 -0.172 0.009 0.011 0.047 -0.371 4.265 0.000
Low 0.152 -0.181 0.011 0.017 0.049 -0.358 3.987 0.003
SOC
High 0.137 -0.167 0.010 0.013 0.044 -0.419 4.434 0.000
Low 0.159 -0.194 0.012 0.020 0.049 -0.475 4.309 0.000
GOV
High 0.142 -0.182 0.010 0.012 0.047 -0.500 4.419 0.000
Low 0.165 -0.188 0.013 0.019 0.049 -0.440 4.315 0.000
ESG
High 0.142 -0.169 0.010 0.011 0.046 -0.427 4.309 0.000
Low 0.157 -0.190 0.013 0.018 0.050 -0.416 4.116 0.001
MKT 0.114 -0.172 0.007 0.012 0.041 -0.693 4.628 0.000
This table presents the descriptive statistics of equally weighted portfolios returns constructed based on the Best-in-Class screen approach. The high (low) portfolios are formed with the 20% high (low) rated companies of each industry according to each ESG score. The maximum, minimum, mean, median, standard deviation, skewness, kurtosis, and the Jarque-Bera probability test of portfolios’ returns for each dimension.
Appendix F – Portfolio performance estimates of the Best-in-Class approach- Carhart (1997) four-factor model
Portfolio α Market SMB HML MOM R2
ENV High 0.002* 1.014*** 0.184*** 0.162*** -0.056*** 0.944
Low 0.003*** 0.986*** 0.455*** 0.215*** -0.030 0.930
Long-Short -0.001 0.028 -0.271*** -0.053 -0.026 0.160
SOC High 0.002*** 0.987*** 0.104*** 0.139*** -0.047*** 0.971
Low 0.004*** 0.996*** 0.492*** 0.151*** -0.007 0.930
Long-Short -0.002* -0.009 -0.387*** -0.012 -0.040* 0.353
GOV High 0.002** 1.017*** 0.225*** 0.112*** -0.072*** 0.958
Low 0.005*** 1.000*** 0.458*** 0.098*** -0.035* 0.944
Long-Short -0.003*** 0.017 -0.232*** 0.014 -0.037 0.180
ESG High 0.002*** 1.005*** 0.123*** 0.117*** -0.066*** 0.962
Low 0.005*** 1.008*** 0.509*** 0.130*** -0.020 0.938
Long-Short -0.003*** -0.003 -0.387*** -0.014 -0.047 0.328
This table presents the estimates of the Carhart (1997) four-factor model (Equation 1) from January 2002 to September 2017 on a monthly basis. The R2s, alphas, and factor loadings concerning market, size, value and momentum are reported. The portfolios are equally weighted. The high (low) portfolios are formed with the 20% high (low) rated companies of each industry according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West(1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
57
Appendix G – Portfolio performance estimates of the Best-in-Class approach- Fama and French (2015) five-factor model
Portfolio α Market SMB HML RMW CMA R2
ENV High 0.001 1.055*** 0.175*** 0.132*** 0.053 0.092 0.943 Low 0.003*** 0.991*** 0.459*** 0.185*** -0.007 -0.091 0.932
Long-Short -0.002* 0.064* -0.284*** -0.053 0.060 0.183** 0.186 SOC
High 0.002*** 1.023*** 0.101*** 0.124*** 0.056 0.062 0.970 Low 0.004*** 0.998*** 0.512*** 0.113 0.030 -0.134* 0.933
Long-Short -0.002** 0.025 -0.411*** 0.011 0.026 0.196*** 0.381 GOV
High 0.001* 1.055*** 0.201*** 0.081** 0.007 0.114** 0.955 Low 0.005*** 1.026*** 0.477*** 0.070* 0.069 -0.115* 0.947
Long-Short -0.003*** 0.030 -0.276*** 0.011 -0.062 0.230*** 0.251 ESG
High 0.001* 1.046*** 0.107*** 0.100*** 0.031 0.099** 0.959 Low 0.005*** 1.015*** 0.529*** 0.096** 0.027 -0.139** 0.942
Long-Short -0.003*** 0.031 -0.422*** 0.004 0.003 0.238*** 0.371
This table presents the estimates of the Fama and French (2015) five-factor model (Equation 2) from January 2002 to September 2017 on a monthly basis. The R2s, alphas, and factor loadings concerning market, size, value, investment and profitability are reported. The portfolios are equally weighted. The high (low) portfolios are formed with the 20% high (low) rated companies of each industry according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West(1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
Appendix H – Portfolio performance estimates depending on the cut-off- Carhart (1997) four-factor model
Portfolio Cut-off
10% 20% 25% 50%
ENV
High 0.002*** 0.002*** 0.002*** 0.003*** Low 0.001 0.002* 0.003*** 0.003***
Long-Short 0.001 0.000 -0.001 0.000 SOC
High 0.001** 0.002*** 0.002*** 0.003*** Low 0.003*** 0.004*** 0.004*** 0.003***
Long-Short -0.002 -0.002 -0.001 -0.001 GOV
High 0.003*** 0.002** 0.002*** 0.002*** Low 0.006*** 0.005*** 0.004*** 0.004***
Long-Short -0.003** -0.003*** -0.003*** -0.002** ESG
High 0.003*** 0.002*** 0.002*** 0.002*** Low 0.005*** 0.004*** 0.004*** 0.004***
Long-Short -0.002 -0.002** -0.002** -0.001*
This table presents the alphas of the Carhart (1997) four-factor model (Equation 1) from January 2002 to September 2017 on a monthly basis. The portfolios are constructed on an equally weighted scheme according to 10%, 20%, 25% and 50% cut-off. The high (low) portfolios are formed with the high (low) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West(1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
58
Appendix I - Portfolio performance estimates depending on the cut-off- Fama and French (2015) five-factor model
Portfolio Cut-off
10% 20% 25% 50%
ENV
High 0.001* 0.001 0.001 0.002*** Low 0.002 0.002* 0.003*** 0.003***
Long-Short 0.000 -0.001 -0.002* -0.001* SOC
High 0.001 0.001* 0.002** 0.002*** Low 0.003*** 0.004*** 0.004*** 0.003***
Long-Short -0.002 -0.002** -0.002** -0.001* GOV
High 0.002*** 0.001 0.001 0.002** Low 0.006*** 0.005*** 0.004*** 0.003***
Long-Short -0.004** -0.004*** -0.003*** -0.002*** ESG
High 0.002*** 0.001* 0.001** 0.002*** Low 0.005*** 0.004*** 0.004*** 0.003***
Long-Short -0.002* -0.003*** -0.003*** -0.002**
This table presents the alphas of the Fama and French (2015) five-factor model (Equation 2) from January 2002 to September 2017 on a monthly basis. The portfolios are constructed on an equally weighted scheme according to 10%, 20%, 25% and 50% cut-off. The high (low) portfolios are formed with the 20% (10%, 25% and 50%) high (low) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West(1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
Appendix J - Portfolio performance estimates for different subperiods- Carhart (1997) four-factor model
Portfolio Subperiod
2002-2006 2007-2011 2012-2017
ENV
High 0.003** 0.003*** -0.000 Low 0.004** 0.002 0.001
Long-Short -0.001 0.001 -0.001 SOC
High 0.003*** 0.004*** 0.000 Low 0.004** 0.004** 0.003
Long-Short -0.001 -0.001 -0.003 GOV
High 0.001 0.003** -0.000 Low 0.006*** 0.005*** 0.004**
Long-Short -0.004*** -0.002 -0.004** ESG
High 0.003** 0.003*** -0.000 Low 0.005** 0.004** 0.003*
Long-Short -0.002 -0.001 -0.003*
This table presents the alphas of the Carhart (1997) four-factor model (Equation 1) for three subperiods between January 2002 to September 2017 on a monthly basis. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The portfolios are equally weighted. Standard errors were estimated using White (1980) or Newey-West(1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
59
Appendix K - Portfolio performance estimates for different subperiods- Fama and French (2015) five-factor model
Portfolio Subperiod
2002-2006 2007-2011 2012-2017
ENV
High 0.002 0.003*** -0.001 Low 0.004** 0.002 0.002
Long-Short -0.002 0.002 -0.003 SOC
High 0.002** 0.003*** -0.001 Low 0.003* 0.004** 0.004**
Long-Short -0.001 -0.001 -0.004** GOV
High 0.001 0.004*** -0.001 Low 0.005*** 0.006*** 0.005***
Long-Short -0.004*** -0.002 -0.006*** ESG
High 0.002* 0.003*** -0.001 Low 0.004** 0.005** 0.004**
Long-Short -0.002 -0.001 -0.005***
This table presents the alphas of the Fama and French (1997) five-factor model (Equation 2) for three subperiods between January 2002 to September 2017 on a monthly basis. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The portfolios are equally weighted. Standard errors were estimated using White (1980) or Newey-West (1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
60
Appendix L - Descriptive statistics of portfolio returns without financial firms
Portfolio Max. Min. Mean Median Std.Dev. Skewness Kurtosis JB Prob.
ENV
High 0.134 -0.175 0.010 0.012 0.044 -0.527 4.458 0.000
Low 0.147 -0.193 0.012 0.016 0.051 -0.474 3.998 0.001
SOC
High 0.111 -0.168 0.009 0.012 0.041 -0.589 4.623 0.000
Low 0.157 -0.186 0.013 0.019 0.052 -0.434 3.916 0.002
GOV
High 0.132 -0.177 0.010 0.012 0.045 -0.504 4.578 0.000
Low 0.177 -0.196 0.014 0.019 0.051 -0.431 4.262 0.000
ESG
High 0.119 -0.165 0.009 0.011 0.042 -0.461 4.457 0.000
Low 0.166 -0.191 0.014 0.016 0.051 -0.405 4.229 0.000
MKT 0.114 -0.172 0.007 0.012 0.041 -0.693 4.628 0.000
This table presents the descriptive statistics of non-financial portfolios returns constructed based on the positive screen approach. The portfolios are constructed on an equally weighted scheme excluding the firms from the financial sector. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The maximum, minimum, mean, median, standard deviation, skewness, kurtosis, and the Jarque-Bera probability test of portfolios’ returns for each dimension.
Appendix M - Performance estimates of portfolios without financial firms– Carhart (1997) four-factor model
Portfolio α Market SMB HML MOM R2
ENV
High 0.002*** 0.995*** 0.120*** 0.044 -0.028 0.951
Low 0.003*** 1.028*** 0.560*** -0.084* -0.027 0.924
Long-Short -0.001 -0.032 -0.440*** 0.128*** -0.001 0.345
SOC
High 0.002*** 0.940*** 0.074** 0.032 -0.007 0.948
Low 0.004*** 1.057*** 0.587*** -0.076* -0.010 0.921
Long-Short -0.002** -0.117*** -0.513*** 0.108* 0.003 0.444
GOV
High 0.002*** 0.981*** 0.162*** 0.065* -0.054*** 0.944
Low 0.006*** 1.046*** 0.512*** -0.038 -0.048* 0.925
Long-Short -0.003*** -0.065** -0.350*** 0.104* -0.006 0.303
ESG
High 0.002*** 0.942*** 0.088*** 0.020 -0.052*** 0.950
Low 0.005*** 1.007*** 0.592*** -0.050 -0.023 0.915
Long-Short -0.003*** -0.065** -0.504*** 0.071 -0.029 0.408
This table presents the estimates of the Carhart (1997) four-factor model (Equation 1) from January 2002 to September 2017 on a monthly basis. The R2s, alphas, and factor loadings concerning market, size, value and momentum are reported. The portfolios are constructed on an equally weighted scheme excluding the firms from the financial sector. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West (1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
61
Appendix N - Performance estimates of portfolios without financial firms– Fama and French (2015) five-factor model
Portfolio α Market SMB HML RMW CMA R2
ENV
High 0.001 1.047*** 0.134*** -0.000 0.147*** 0.104** 0.954
Low 0.003*** 1.024*** 0.564*** -0.122** -0.034 -0.119 0.927
Long-Short -0.002** 0.023 -0.429*** 0.122** 0.181*** 0.224*** 0.399
SOC
High 0.001 0.990*** 0.093*** -0.023 0.174*** 0.131*** 0.954
Low 0.005*** 1.053*** 0.610*** -0.110** 0.013 -0.195** 0.927
Long-Short -0.003*** -0.063* -0.517*** 0.088* 0.162** 0.326*** 0.510
GOV
High 0.001 1.031*** 0.164*** 0.030 0.096* 0.105 0.944
Low 0.005*** 1.079*** 0.530*** -0.070 0.078 -0.115 0.927
Long-Short -0.004*** -0.048 -0.366*** 0.099** 0.018 0.220*** 0.338
ESG
High 0.001 1.004*** 0.090*** -0.025 0.137*** 0.174*** 0.953
Low 0.005*** 1.033*** 0.627*** -0.102** 0.102 -0.136* 0.922
Long-Short -0.004*** -0.028 -0.537*** 0.076 0.034 0.311*** 0.465
This table presents the estimates of the Fama and French (2015) five-factor model (Equation 2) from January 2002 to September 2017 on a monthly basis. The R2s, alphas, and factor loadings concerning market, size, value, investment and profitability are reported. The portfolios are constructed on an equally weighted scheme excluding the firms from the financial sector. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West (1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
62
Appendix O – Portfolio performance estimates of the value weighted scheme – Carhart (1997) four-factor model with dummies
Portfolio α Market SMB HML MOM
R2
EX Dt EX Dt EX Dt EX Dt EX Dt
ENV
High 0.002*** 0.009*** 0.965*** -0.097** -0.221*** -0.092 -0.001 0.123* -0.033 0.066** 0.956
Low 0.005*** 0.005** 0.992*** -0.163*** 0.156*** -0.113 0.042 0.035 -0.002 -0.067 0.907
Long-Short -0.002 0.004 -0.027 0.066 -0.377*** 0.020 -0.043 0.087 -0.032 0.133** 0.261
SOC
High 0.003*** 0.005 0.931*** -0.059 -0.245*** 0.115 -0.024 0.161*** -0.026 0.082** 0.951
Low 0.007*** -0.001 0.990*** -0.075 0.279*** 0.199 0.012 -0.104 0.064** -0.135*** 0.908
Long-Short -0.004*** 0.005 -0.059* 0.015 -0.524*** -0.084 -0.036 0.266** -0.089** 0.217*** 0.438
GOV
High 0.003*** 0.005 0.958*** -0.069 -0.165*** -0.019 -0.043 0.070 -0.036 0.113*** 0.945
Low 0.007*** -0.001 1.010*** -0.085 0.142** 0.092 -0.050 -0.088 0.018 -0.111** 0.899
Long-Short -0.004*** 0.006 -0.052* 0.016 -0.307*** -0.111 0.007 0.158 -0.054 0.224*** 0.258
ESG
High 0.002*** 0.008*** 0.931*** -0.077* -0.233*** -0.075 -0.038 0.186*** -0.032 0.107*** 0.950
Low 0.009*** 0.001 0.974*** -0.155* 0.277*** -0.003 0.048 -0.046 0.097 -0.192*** 0.886
Long-Short -0.005*** 0.008 -0.045 0.081 -0.512*** -0.069 -0.085 0.243* -0.129 0.304*** 0.386
This table presents the results of the Carhart (1997) four-factor model with dummies (Equation 3) from January 2002 to June 2017 on a monthly basis. The estimates correspond to the Best-in-Class ESG portfolios. The R2s, alphas, and factor loadings concerning market, size, value and momentum are reported for periods of expansion
(Ex) and recession (Dt). Periods of recession and expansion were defined using NBER business cycles. The high (low) portfolios are formed with the 20% high (low) rated
companies according to each ESG score. The portfolios are equally weighted. Standard errors in parentheses were estimated using White (1980) or Newey-West (1987). ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
63
Appendix P - Portfolio performance estimates of the value weighted scheme – Fama and French (2015) five-factor model with dummies
Portfolio α Market SMB HML RMW CMA
R2
EX Dt EX Dt EX Dt EX Dt EX Dt EX Dt
ENV
High 0.002** 0.011*** 0.986*** -0.089** -0.234*** -0.120 -0.017 0.122 0.045 -0.126 0.153*** 0.194 0,957
Low 0.005*** 0.008** -0.234*** -0.134* 0.152*** -0.095 0.036 0.085 -0.053 -0.097 -0.032 -0.016 0,906
Long-Short -0.003** 0.003 -0.017 0.044 -0.386*** -0.025 -0.053 0.036 0.098 -0.029 0.185 0.210 0,277
SOC
High 0.002*** 0.007*** 0.045 -0.125*** -0.252*** 0.053 -0.051 0.136* 0.081** -0.407*** 0.179*** 0.154 0,956
Low 0.007*** -0.002 0.153*** -0.064 0.324*** 0.192 0.014 -0.076 0.049 0.136 -0.183** -0.270 0,910
Long-Short -0.005*** 0.009* 0.011*** -0.061 -0.575*** -0.139 -0.065 0.212* 0.031 -0.543* 0.361*** 0.423 0,478
GOV
High 0.002*** 0.005** -0.089** -0.071 -0.186*** -0.055 -0.090 0.051 0.033 0.005 0.232*** 0.108 0,949
Low 0.007*** -0.004* -0.120 -0.121 0.161*** 0.202 -0.027 0.030 0.012 0.274 -0.144* -0.709** 0,902
Long-Short -0.005*** 0.008** 0.122 0.050 -0.347*** -0.257 -0.063 0.021 0.021 -0.268 0.376*** 0.817* 0,313
ESG
High 0.002** 0.009*** -0.126 -0.105** -0.247*** -0.117 -0.071 0.178** 0.067 -0.138 0.203*** 0.139 0,955
Low 0.009*** -0.002 0.194 -0.145* 0.344*** 0.035 0.048 0.063 0.084 0.222 -0.216 -0.474 0,889
Long-Short -0.006*** 0.010* 0.002** 0.042 -0.593*** -0.149 -0.116 0.123 -0.020 -0.337 0.416** 0.609 0,418
This table presents the results of Fama and French (2015) five-factor model with dummies (Equation 4) from January 2002 to June 2017 on a monthly basis. The estimates correspond to the Best-in-Class ESG portfolios. The R2s, alphas, and factor loadings concerning market, size, value, investment and profitability are reported for periods of
expansion (Ex) and recession (Dt). Periods of recession and expansion were defined using NBER business cycles. The high (low) portfolios are formed with the 20% high (low)
rated companies according to each ESG score. The portfolios are equally weighted. Standard errors in parentheses were estimated using White (1980) or Newey-West (1987). ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
64
Appendix Q - Portfolio performance estimates of the Best-in-Class screen approach – Carhart (1997) four-factor model with dummies
Portfolio α Market SMB HML MOM
R2
EX Dt EX Dt EX Dt EX Dt EX Dt
ENV
High 0.001 0.004** 1.036*** -0.038 0.185*** -0.187** 0.173*** -0.010 -0.043 -0.033 0,946
Low 0.002** -0.000 1.025*** -0.116* 0.454*** -0.293* 0.290*** -0.259*** 0.002 -0.139*** 0.938
Long-Short -0.001 0.004 0.011 0.078 -0.269*** 0.106 -0.117** 0.249** -0.044 0.106* 0.198
SOC
High 0.002** 0.006** 1.005*** -0.009 0.105*** -0.181* 0.150*** -0.010 -0.034* -0.022 0.973
Low 0.003*** 0.002 1.007*** 0.018 0.481*** -0.163 0.231*** -0.329*** 0.046 -0.151*** 0.937
Long-Short -0.002* 0.004 -0.002 -0.027 -0.377*** -0.019 -0.081* 0.319*** -0.080*** 0.129** 0.409
GOV
High 0.001* 0.001 1.032*** -0.025 0.211*** 0.043 0.160*** -0.166** -0.062*** -0.035 0.960
Low 0.005*** -0.004 1.007*** -0.032 0.452*** -0.099 0.156*** -0.260*** 0.012 -0.158*** 0.950
Long-Short -0.003*** 0.005 0.024 0.007 -0.241*** 0.142 0.004 0.094 -0.074* 0.123** 0.224
ESG
High 0.001* 0.005* 1.030*** -0.045 0.122*** -0.160 0.134*** -0.024 -0.061*** -0.014 0.964
Low 0.004*** 0.002 1.035*** -0.049 0.497*** -0.143 0.209*** -0.286*** 0.015 -0.118** 0.943
Long-Short -0.003** 0.003 -0.005 0.004 -0.375*** -0.016 -0.075 0.263** -0.077* 0.104** 0.361
This table presents the results of the Carhart (1997) four-factor model with dummies (Equation 3) from January 2002 to June 2017 on a monthly basis. The estimates correspond to the Best-in-Class ESG portfolios. The R2s, alphas, and factor loadings concerning market, size, value and momentum are reported for periods of expansion (Ex)
and recession (Dt). Periods of recession and expansion were defined using NBER business cycles. The portfolios are equally weighted. The high (low) portfolios are formed
with the 20% high (low) rated companies of each industry according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West(1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
65
Appendix R - Portfolio performance estimates of the Best-in-Class screen approach – Fama and French (2015) five-factor model with dummies
Portfolio α Market SMB HML RMW CMA
R2 EX Dt EX Dt EX Dt EX Dt EX Dt EX Dt
ENV High 0.000 0.009*** 1.066*** -0.001 0.181*** -0.204* 0.122** 0.073 0.063 -0.163 0.106 0.018 0,9451
Low 0.002** 0.003 1.025*** 0.013 0.472*** -0.279* 0.292*** -0.208* -0.013 0.179 -0.184** 0.146 0.938
Long-Short -0.002* 0.006 0.040 -0.014 -0.291*** 0.075 -0.170*** 0.281** 0.075 -0.342 0.290*** -0.128 0.248
SOC High 0.001 0.009*** 1.031*** 0.024 0.103*** -0.189* 0.116*** 0.058 0.058* -0.075 0.076* 0.009 0.972
Low 0.004*** 0.001 0.994*** 0.132 0.515*** -0.153 0.214*** -0.272** 0.007 0.372 -0.178** -0.086 0.939
Long-Short -0.003** 0.008** 0.037 -0.108 -0.412*** -0.035 -0.097* 0.330*** 0.052 -0.447* 0.254*** 0.095 0.447
GOV High 0.001 0.007* 1.056*** -0.019 0.190*** 0.050 0.103** -0.091 0.017 -0.273 0.140* -0.118 0,9578
Low 0.005*** -0.000 1.022*** 0.091 0.485*** -0.114 0.140*** -0.219** 0.060 0.111 -0.155** 0.124 0.950
Long-Short -0.004*** 0.007* 0.035 -0.110 -0.295*** 0.164 -0.037 0.128 -0.043 -0.384* 0.295*** -0.243 0.304
ESG High 0.001 0.010*** 1.060*** 0.004 0.107*** -0.172* 0.100** 0.035 0.036 -0.150 0.102 0.104 0.962
Low 0.004*** 0.002 1.037*** 0.028 0.529*** -0.129 0.207*** -0.252** 0.014 0.256 -0.209*** -0.048 0.947
Long-Short -0.003*** 0.008** 0.023 -0.025 -0.422*** -0.043 -0.107* 0.287*** 0.022 -0.406 0.311*** 0.151 0.427
This table presents the results of the Fama and French (2015) five-factor model with dummies (Equation 4) from January 2002 to June 2017 on a monthly basis. The estimates correspond to the Best-in-Class ESG portfolios. The R2s, alphas, and factor loadings concerning market, size, value, investment and profitability are reported for periods of
expansion (Ex) and recession (Dt). Periods of recession and expansion were defined using NBER business cycles. The portfolios are equally weighted. The high (low) portfolios
are formed with the 20% high (low) rated companies of each industry according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West (1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
66
Appendix S - Portfolio performance estimates depending on the cut-off– Carhart (1997) four-factor model with dummies
Portfolio
Cut-off
10% 20% 25% 50%
EX Dt EX Dt EX Dt EX Dt
ENV
High 0.002* 0.005 0.001 0.005* 0.001* 0.003 0.002*** 0.001
Low 0.001 -0.005 0.001 -0.000 0.002** 0.001 0.003*** 0.000
Long-Short 0.001 0.010 -0.000 0.005 -0.001 0.002 0.000 0.001
SOC
High 0.001 0.007** 0.002** 0.004* 0.002*** 0.004* 0.002*** 0.002
Low 0.003** 0.003 0.003*** -0.001 0.003*** 0.000 0.003*** -0.001
Long-Short -0.002 0.004 -0.002 0.006 -0.001 0.005 -0.001 0.003
GOV
High 0.003*** 0.001 0.001* 0.002 0.001** 0.002 0.002*** 0.002
Low 0.006*** -0.006 0.005*** -0.005 0.004*** -0.005 0.003*** -0.001
Long-Short -0.003** 0.007 -0.004*** 0.007*** -0.003*** 0.007* -0.002** 0.003
ESG
High 0.003*** 0.002 0.002** 0.005* 0.002** 0.005** 0.002*** 0.002
Low 0.004*** -0.002 0.004*** -0.001 0.004*** -0.001 0.003*** -0.002
Long-Short -0.002 0.004 -0.002** 0.006 -0.002** 0.006 -0.001 0.004
This table presents the alphas of the Carhart (1997) four-factor model with dummies (Equation 3) from January 2002 to September 2017 on a monthly basis. The alphas of
High, Low, and Long-Short portfolios are reported for periods of expansion (Ex) and recession (Dt). Periods of recession and expansion were defined using NBER business
cycles. The portfolios are constructed on an equally weighted scheme according to 10%, 20%, 25% and 50% cut-off. The high (low) portfolios are formed with the 20% (10%, 25% and 50%) high (low) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West(1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
67
Appendix T - Portfolio performance estimates depending on the cut-off – Fama and French (2015) five-factor model with dummies
Portfolio
Cut-off
10% 20% 25% 50%
EX Dt EX Dt EX Dt EX Dt
ENV
High 0.001 0.010*** 0.000 0.009*** 0.0006 0.007** 0.002*** 0.005
Low 0.001 0.007 0.002 0.007 0.002** 0.006 0.003*** 0.002
Long-Short 0.000 0.002 -0.001 0.002 -0.002* 0.001 -0.001 0.003
SOC
High 0.000 0.008** 0.001 0.007** 0.001 0.007*** 0.002** 0.006**
Low 0.003** 0.002 0.003*** -0.002 0.003*** -0.002 0.003*** 0.001
Long-Short -0.003* 0.005 -0.003** 0.008* -0.002** 0.009** -0.001* 0.006*
GOV
High 0.002** 0.002 0.001 0.007** 0.001 0.006* 0.001** 0.004
Low 0.006*** -0.003 0.005*** 0.000 0.004*** 0.001 0.003*** 0.003
Long-Short -0.004*** 0.006 -0.005*** 0.007 -0.004*** 0.006 -0.002*** 0.001
ESG
High 0.002** 0.006* 0.001 0.007** 0.001 0.009*** 0.001** 0.006**
Low 0.005*** 0.002 0.004*** 0.003 0.004*** 0.002 0.003*** 0.001
Long-Short -0.003** 0.004 -0.003*** 0.005 -0.003*** 0.007 -0.002** 0.005*
This table presents the alphas of the Fama and French (2015) five-factor model with dummies (Equation 4) from January 2002 to September 2017 on a monthly basis. The
alphas of High, Low, and Long-Short portfolios are reported for periods of expansion (Ex) and recession (Dt). Periods of recession and expansion were defined using NBER
business cycles. The portfolios are constructed on an equally weighted scheme according to 10%, 20%, 25% and 50% cut-off. The high (low) portfolios are formed with the 20% (10%, 25% and 50%) high (low) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West(1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
68
Appendix U – Performance estimates of portfolios without financial firms– Carhart (1997) four-factor model with dummies
Portfolio α Market SMB HML MOM
R2
EX Dt EX Dt EX Dt EX Dt EX Dt
ENV High 0.001* 0.004 1.009*** 0.001 0.099*** 0.138 0.082** -0.112 -0.028 0.016 0.953
Low 0.003** 0.001 1.056*** -0.074 0.556*** -0.199 -0.030 -0.194* 0.004 -0.115** 0.928
Long-Short -0.001 0.003 -0.047 0.074 -0.458*** 0.338* 0.112** 0.081 -0.033 0.131** 0.377
SOC High 0.002** 0.005* 0.943*** 0.051 0.056* 0.129 0.062* -0.092 -0.010 0.038 0.951
Low 0.003*** 0.003 1.087*** -0.035 0.575*** -0.208 0.020 -0.345*** 0.026 -0.122** 0.928
Long-Short -0.002 0.002 -0.144*** 0.086 -0.519*** 0.337* 0.042 0.253** -0.036 0.160*** 0.484
GOV High 0.002** 0.001 0.987*** 0.018 0.146*** 0.077 0.128*** -0.226*** -0.054* -0.010 0,947
Low 0.005*** -0.004 1.056*** -0.033 0.500*** -0.117 0.052 -0.396*** 0.017 -0.215*** 0.935
Long-Short -0.003*** 0.005 -0.070** 0.051 -0.353*** 0.194 0.076 0.170* -0.071** 0.206*** 0.362
ESG High 0.002** 0.004 0.953*** 0.015 0.069** 0.113 0.057 -0.109 -0.052** 0.019 0.952
Low 0.004*** 0.002 1.040*** -0.058 0.580*** -0.273 0.041 -0.351*** 0.043 -0.204*** 0.926
Long-Short -0.003** 0.002 -0.087** 0.073 -0.512*** 0.386** 0.016 0.243*** -0.095** 0.223*** 0,468
This table presents the estimates of the Carhart (1997) four-factor model with dummies (Equation 3) from January 2002 to September 2017 on a monthly basis. The R2s,
alphas, and factor loadings concerning market, size, value and momentum are reported for periods of expansion (Ex) and recession (Dt). Periods of recession and expansion
were defined using NBER business cycles. The portfolios are constructed on an equally weighted scheme excluding the firms from the financial sector. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West (1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
69
Appendix V – Performance estimates of portfolios without financial firms– Fama and French (2015) five-factor model with dummies
Portfolio α Market SMB HML RMW CMA
R2 EX EX EX EX EX EX
ENV High 0.001 0.005 1.055*** -0.040 0.117*** 0.113 0.028 -0.099 0.151*** -0.185 0.112** -0.078 0.956
Low 0.003*** 0.002 1.048*** 0.002 0.568*** -0.160 -0.041 -0.158 -0.040 0.163 -0.179** -0.007 0.930
Long-Short -0.002** 0.003 0.007 -0.042 -0.451*** 0.273 0.069 0.059 0.190*** -0.349 0.291*** -0.071 0.426
SOC
High 0.001 0.005 0.989*** -0.017 0.081** 0.076 -0.008 -0.064 0.180*** -0.167 0.157*** -0.174 0.957
Low 0.004*** 0.000 1.080*** 0.041 0.611*** -0.174 0.031 -0.294** -0.010 0.499* -0.282*** -0.224 0.936
Long-Short -0.003*** 0.005 -0.092** -0.058 -0.530*** 0.250 -0.039 0.230* 0.190*** -0.667** 0.439*** 0.050 0.565
GOV
High 0.001 0.005 1.028*** 0.029 0.152*** 0.059 0.080 -0.197*** 0.097 -0.186 0.108 0.014 0,947
Low 0.005*** 0.002 1.071*** 0.118 0.537*** -0.130 0.029 -0.324*** 0.071 0.022 -0.156* 0.145 0.932
Long-Short -0.004*** 0.003 -0.043 -0.089 -0.385*** 0.189 0.051 0.127 0.027 -0.208 0.264*** -0.131 0.357
ESG
High 0.000 0.007** 1.002*** 0.004 0.072** 0.085 -0.016 -0.066 0.138*** -0.189 0.204*** -0.079 0.956
Low 0.004*** 0.004 1.050*** 0.083 0.637*** -0.276 0.014 -0.269** 0.090 0.234 -0.207** 0.045 0.929
Long-Short -0.004*** 0.003 -0.048 -0.079 -0.565*** 0.361** -0.030 0.203 0.048 -0.423 0.412*** -0.124 0.515
This table presents the estimates of the Fama and French (2015) five-factor model with dummies (Equation 4) from January 2002 to September 2017 on a monthly basis. The
R2s, alphas, and factor loadings concerning market, size, value, investment and profitability are reported for periods of expansion (Ex) and recession (Dt). Periods of recession
and expansion were defined using NBER business cycles. The portfolios are constructed on an equally weighted scheme excluding the firms from the financial sector. The high (low) portfolios are formed with the 20% high (low) rated companies according to each ESG score. The Long-Short portfolio trades the High portfolio long while the Low portfolio is traded short. Standard errors were estimated using White (1980) or Newey-West (1987) adjustment. ∗∗∗, ∗∗, and ∗ indicates significance at the 1%, 5%, and 10% level.
70
