(ir)responsible investing: revisiting the effects of esg
TRANSCRIPT
(Ir)responsible Investing: Revisiting the Effects of ESG Performance on Portfolio Returns.
Master Thesis
MSc. Finance
January 11th, 2021 Author: N.P.A. van Tilborg Email: [email protected] Student ID: S2305534 Supervisor: Dr. A. Dalò
2
Abstract
The sin stock anomaly where stocks belonging to “sin” industries such as alcohol, tobacco,
gambling and weapons yield abnormal returns above the market return was observed decades
ago. Presently, this anomaly fits into the broader focus of investors on factors besides returns,
such as ESG, and indications exist that high ESG might result in excess returns. In this paper
we investigate if ESG and returns are related. In order to investigate this, we create decile
portfolios based on the Refinitiv ESG Combined Scores of the individual stocks and test for the
presence of alpha in these portfolios using different models in the period 2003-2019.
Throughout the models and after robustness checks, it seems that actually the lowest ESG
portfolio is the only one to yield a significantly positive alpha.
3
Contents
1. Introduction 3
2. Literature review 4
3. Methodology 6
4. Data 9
5. Results 13
6. Robustness 19
7. Conclusion 19
8. Discussion 20
9. Appendix 21
10. References 23
4
1. Introduction
Investors face increasing pressure to carefully select their stocks. This is especially true
for institutional investors such as pension funds whom besides being expected to deliver good
returns are expected to behave ethically and select their portfolio accordingly. This increasing
interest in socially responsible investing (SRI) has pushed (institutional) investors to focus on
other preferences beyond returns such as performance on environmental, social and governance
(ESG) issues.
The increased interest in ESG investing becomes apparent from the large inflow of
capital. According to the Global Sustainable Investment Alliance (GSIA), a collaboration
between seven (inter)nation sustainable investment organisations, the global market for
sustainable investment assets is valued at $30.7 trillion at the beginning of the year, an increase
of 34% in just two years. (GSIA, 2018)
This growth is still ongoing as the Financial times reported. In 2020 in the period
January 1st until July 30th ETFs focusing on ESG pulled in $38bn, a record, in new capital.
(Nauman, 2020)
The debate on the responsibility of investors for the effects their investments has been
going on for decades, where previously the focus was mainly on the question if investors should
avoid investing in stocks that had negative effects on society such as tobacco, alcohol, gambling
and weapons collectively known as “sin stocks” or “vice stocks”. A commonly used argument
against excluding these stocks from a portfolio has long been the high returns these sin stocks
offer compared to the market, which has been dubbed the sin premium. The excess return of
sin stocks has been extensively research in the past, and it is generally accepted that these stocks
do offer excess returns.
On the other hand, research into returns of stocks with high ESG scores has recently
become increasingly popular, as scholars suggest that besides avoiding sin stocks, investors
should invest in stocks that have an above average ESG score. Thus far, results into the
performance have been ambiguous with some suggesting these high ESG stocks yield
significantly better results while others find the opposite, besides some studies find no
significant effects be it positive or negative. In the literature review, these studies are explored
more thoroughly.
5
In this paper we use Fama and French’s (2015) five factor model to to test for the
presence of alpha in order to investigate if there is a link between ESG performance and
portfolio returns.
Does ESG performance affect portfolio returns?
In the next section we will examine the literature that has already been published on this
subject. In section 3 the methodology and model will be explained, while in section 4 an
overview of the data will be presented. In section 5 the results of the analysis will be presented,
while in section 6 we will check the robustness of our results. a conclusion from these results
is drawn in section 7 followed by a discussion in section 8.
2. Literature Review. As mentioned, the relationship between ESG and performance was first studied in the
context of abnormal returns for sin stocks. One of the first to observe and give an explanation
for the sin stock anomaly was Merton (1987). He argues that the excess returns on sin stocks
can be explained by their susceptibility to being mispriced, since there are few analysts reports
on these stocks because many investors are avoiding these stocks resulting in a lower demand.
Another factor that can explain the anomaly is the high(er) risk of litigation already being
factored into the price. The issue of mispricing also surfaces in a study by Mǎnescu (2011), in
which she concludes that there is an effect of ESG factors on stock return but that these effects
are not sufficiently incorporated in stock prices.The sin stock anomaly was also observed by
Salaber (2007) who contributed to the field by finding that in Europe, the excess returns of sin
stocks are larger in traditionally protestant countries when compared to catholic nations. This
is explained by a larger “sin aversion” being present in protestants, resulting in them demanding
higher returns as a form of compensation for holding on to these stocks.
The sin stock anomaly is not just limited to European markets. Hong and Kacperczyk
(2009) found a similar anomaly when they focused on US sin stocks. They created an equally
weighted portfolio of sin stocks and used both Fama and French’s (1993) three-factor model
and the Carhart (1997) four-factor model to evaluate returns and compare them to a market
benchmark. They concluded that the higher returns for sin stocks can be mostly attributed to
the lack of interest from large institutional investors.
In terms of stocks with high ESG scores there is some evidence that these stocks also
provide an excess return. For example, Sahut and Pasquini-Descomps (2015) found that the
overall ESG score can have a significant positive effect on returns in the UK, though they did
6
not find similar evidence for the US and Switzerland which is exemplary for the ambiguous
results when studying the relationship between ESG performance and returns.
This is also exemplified in a study by Doreitner Utz and Wimmer (2013) that found a
significant positive relationship between ESG performance and returns in the long run on both
the European and the North American market. However, in the same study the effect of ESG
on returns is much weaker in Japan.
La Torre et al. (2020) examine the relationship between ESG scores and returns of
stocks listed on the Eurostoxx50 index. They found some weak results, for some of the stocks
listed on the Eurostoxx50 index, there is a small significant relation between ESG and returns.
For the majority of the firms however, this was not the case.
Despite the previous literature almost unanimously suggesting that sin stocks yield
excess returns there is some recent literature that seems to reject this and find that ESG
performance does not affect returns. Richey (2017) analysed US sin stocks and confirmed the
results of most previous studies in that under the CAPM, the three- and four-factor models, sin
stocks seem to provide a significant positive alpha indicating excess returns for the sin stocks.
However, when using the Fama French (2015) five-factor model by including the investment
and the profitability factor, the alpha becomes insignificant suggesting there is no abnormal
return on sin stocks when controlling for these two additional factors.
Another study that could not establish a link between low ESG stocks and excess returns
was by Hoeper and Zeume (2014) who analysed the performance of the Vice Fund, an
investment funds based entirely on sin stocks. They did not find abnormal returns for the Vice
Fund. Which they link to literature regarding ethical funds that also fail to establish a connection
between fund ethics and returns. This would imply that funds regardless of their ESG
characteristics are generally priced correctly and thus offer no opportunity to outperform the
market since any benefits from the stock selection are already priced into the fund.
This conclusion is also interesting when used in a different context, Humphrey et al.
(2012) investigated the consequences of implementing corporate social performance (CSP)
strategies on financial performance and risk. They analysed a set of UK firms with differing
ESG characteristics and found that for their sample there was no significant effect of ESG on
performance or risk, implying that implementing a CSP strategy does not give a significant
benefit nor cost in terms of returns.
7
In a study that compares performance of ESG indices against the performance regular
MSCI indices by Jain, Deep Sharma and Srivastava (2019) found no evidence that the indices
based on high ESG scores yield significantly different returns compared to the MSCI indices.
Besides evidence suggesting that ESG scores have a positive or neutral effect on returns,
there is also evidence that high ESG levels might negatively impact stock performance.
Chawana (2014) analysed the South African market and specifically the returns of a
dedicated SRI index. When comparing the performance of this index to other indices it seems
that investors actually pay a premium for their SRI investments.
In addition, Auer and Schuhmacher (2016) found that whereas investors focussing on
high ESG scores are able to achieve similar returns to regular portfolios in North America and
in the Asia and Pacific region, European investors seem to be paying a price for their responsible
investments.
In a study by Friede et al. (2015) that examined the existing literature on the link
between ESG and financial performance they addressed the varying conclusions of different
studies. When aggregating the results, they found however that, overall, studies suggest that
there is a positive relation between ESG and financial performance. They do not however that
this relationship is more evident for individual firms than portfolio studies.
Given the results obtained in the existing literature we expect that ESG performance
does affect returns. We therefore hypothesise that ESG does have an effect on portfolio returns.
3. Methodology
The models used in the studies mentioned in the previous section have been updated and
extended. Given the ambiguous results in studies into the link between high ESG scores and
returns, and the slightly older studies claiming excess returns for sin stocks, in this study we
will revisit these outcomes using a more modern model as introduced by Fama and French
(2015). This five-factor model is an extension of the Fama French (1993) three-factor model
and the Carhart (1997) four-factor model that has often been used in previous literature to
explain anomalies in returns.
One of the first and most common models used to determine required returns is the
capital asset pricing model (CAPM). This model finds its roots in modern portfolio theory as
devised by Markowitz (1952), where a link is established between (portfolio) risk and expected
8
returns. This implies that for a given level of risk the expected returns are maximised, and in
order to increase those returns, an investor needs to take on more risk. This model was used as
the basis for William Sharpe (1964) to create the capital asset pricing model. The CAPM adds
to modern portfolio theory by distinguishing between systematic and non-systematic (i.e.
specific) risk. These two types of risk affect investors differently since with proper
diversification it is possible to eliminate specific risk while systematic risk is unavoidable. The
CAPM is expressed as:
!!,#$ = #! + %!!%$ + &&
Here !!,#$ represents portfolio p’s excess returns in a given period t which is defined as
the portfolio return minus the risk-free rate, where #! denotes the risk adjusted returns also
referred to as the pricing anomaly. The portfolio beta is indicated as %! while the market risk
premium (MRP) is included as !%$ .
The risk adjusted return is used to measure a portfolio’s performance compared to the
market. Under the efficient market hypothesis (EMH) alpha would equal zero since with all
available information already reflected in prices, it will not be possible to systematically
outperform the market. The beta measures the systemic risk of a stock or a portfolio compared
to that of the market, where the market beta equals one implies similar risk as the market and a
beta greater than one implies a higher risk compared to the market portfolio, whereas the
opposite is true for a beta smaller than one.
Even though the CAPM is one of the most commonly taught and used models, it does
come with some limitations. In practice it is often inaccurate when predicting returns in
comparison to realised returns. Due to these shortcomings, scholars have worked on improving
the CAPM’s accuracy by including more factors. One of the most used of these models was
developed by Fama and French (1993) who created a three-factor model (FF3) by adding the
firm size and the book to value ratio to the CAPM. In this model firm size is expressed in the
difference in returns between small and large firms. This factor is included to account for the
more volatile nature of small firms compared to larger firms resulting in a higher upside
potential for smaller firms. This factor is known as the “Small-Minus-Big” or SMB factor of
the model. The final factor is known as the “High-Minus-Low” (HML) and is included to
account for the fact that returns for high book-to-market value stocks differ from those of low
9
book-to-market value stock (also known as growth stocks) When adding the size factor and the
value premium to the CAPM we obtain the Fama and French three-factor model:
!!,#$ = #! + %!!%$ + %'()'() + %*(+*(+ + &&
Here SMB represents the size factor where the %'() is the corresponding factor-beta.
HML is the size factor which has a factor-beta presented as %*(+.
Even though the Fama-French three-factor model is more accurate in predicting returns
than the CAPM, it still is not always able to fully explain portfolio returns. Research into
additional factors that might explain the returns of stocks or portfolios continued and this
eventually led to Carhart (1997) to further expand the Fama-French three-factor model to
include a fourth factor, the momentum. This factor accounts for difference in past performance
of stocks. Carhart’s study builds on the assumption that it is possible for investors to benefit
from trends, encouraging investors to buy stocks that have performed well over the last months,
while selling stocks that are down on performance. The momentum factor is derived by
subtracting the excess return of stocks that went down from the excess returns of stocks whose
value went up. This gives the following model:
!!,#$ = #! + %!!%$ + %'()'() + %*(+*(+ + %(,((,( + &&
The additional momentum factor is expressed as MOM which has a corresponding
coefficient expressed as %(,(. The latest contribution to the multi-factor model is again by
Fama and French (2015) who expanded it to become the five-factor model. They included two
“quality factors”. Firstly the profitability factor which is included on the basis that more
profitable firms tend to realise higher returns. Additionally the model includes the investment
factor that factors in the amount of earnings which are reinvested, which should lead to higher
returns. Unlike the Carhart four-factor model, the Fama-French five-factor model does not
include a momentum factor. Ultimately, this leads to the following model:
!!,#$ = #! + %! × !%$ + %'()'() + %*(+*(+ + %-(./(0 + %/(01(2 + &&
10
Here RMW (robust minus weak) is the expression of the profitability factor with %-(.
being the corresponding beta. The investment factor is expressed as Conservative Minus
Aggressive (CMA) with the corresponding beta depicted as %/(0.
While these models help to measure the performance of a given portfolio, in order to
test the hypothesis, we first need to create portfolios that are based on the ESG performance of
the stocks comprising the portfolio.
4. Data
To measure the ESG performance of stock we have chosen to use the Refinitiv ESG
Combined Scores. This score measures a firms ESG performance over the three pillars of ESG:
Environmental, Social and Governance. Each of these pillars consists of different topics. In
total over 450 ESG measures are collected and combined into scores for each individual pillar
and an overall ESG Score. These scores however do not take into account the attention
generated by scandals a company might be involved in. This effect is measured in the ESG
Controversies Score. Combining the ESG and ESG Controversies Scores results in the
Combined ESG Score which is equal to the ESG Score if there are no controversies and is lower
than the ESG Score when such controversies do arise.
To construct the portfolios, using DataStream, we first obtain the market capitalisation
expressed in US dollars and the ESG Combined Scores with all of the components used in
deriving this score (ESG Score, Controversies, Environmental Pilar, Social Pilar and
Governance Pilar) for all stocks that were scored in the period June 30th 2003 until June 30th
2019, where we obtain these scores on a yearly basis. Based on the ESG Combined Score at the
end of June for each year we create ten portfolios based on deciles, where portfolio 1 contains
the 10% of stocks with the highest ESG Combined Scores, portfolio 2 the second best 10%, etc.
Subsequently, for all the stocks included in the portfolio we obtain the prices at the end of each
month included in our sample period, and using these prices we calculate the monthly returns
for each individual stock using:
!1,# = 341,#41,#23
5 − 1
Where !1,# is the return of stock it time t, 41,# refers to the price of stock i at time t.
To prevent large outliers affecting the results of our regressions we chose to trim the
dataset to the 99th percentile, excluding all returns that fall outside of this range.
11
The available data differs considerably for each year with the number of available firms
ranging from 867 firms in 2003 up to 6,259 different firms in 2018. A complete overview of
the number of firms available each year can be found in appendix A.
In order to prevent small cap stocks to disproportionally affect the portfolio returns, all
the stocks are value weighted based on their market capitalisation each year at the end of June,
where the weight of a single stock is equal to the market capitalisation of that stock, divided by
the sum of market capitalisations of all stocks included in the portfolio. Using the weighted
returns, we can then calculate the monthly portfolio returns for each of the portfolios. The
portfolios are rebalanced each year at the end of June to encapsulate the changes in ESG
performance and changes in (relative) market capitalisation. Value weighting also better reflects
typical investors behaviour in the sense that investors tend to invest in firms they are familiar
with which typically have a large market capitalisation.
To test if high ESG portfolios do outperform low ESG portfolios, we create an eleventh
portfolio which consists of a long position in the highest scoring portfolio and a short position
in the lowest ESG scoring portfolio. If this were to be the case, this eleventh portfolio should
give a significantly positive alpha.
4.1 Data overview
Table 1 presents a general description of the dataset. The data appears to be slightly
unbalanced given the different number of observations for the different ESG scoring
components. However, this should not affect the results of our research since for the
construction of our portfolios, only the ESG combined score is used.
12
Table 1:
Descriptive Statistics Raw Data July 2003- June 2020
VARIABLES Mean SD Min Max Obs. '89:;/<8=!>(@9>8ℎBC) 0.008 1.101 -0,690 1.136 1,597,256
(D!;<81D4(E'F@B>. ) 6.801 47.591 0.000 11,389.304 127,116
I'J19@KL><M Score 39.094 19.144 0.000 93.610 68,933
I'J19>!9N<!OL<O Score 91.927 21.534 0.000 100.000 68,912
I'J':9!< 40.398 20.288 0.000 95.070 68,933
I>NL!9@<>8DBPLBBD! Score 30.499 28.671 0.000 99.100 68,792
'9:LDBPLBBD! Score 40.775 23.063 0.052 98.920 68,792
J9N<!>D>:<PLBBD! Score 47.439 22.692 0.140 99.380 68,925
WhereStockreturnsaredefinedas(3!,#/3!,#&')-1,MarketcapisthevalueofoutstandingstocksonJune
30thexpressedinUSDollars.AllscoresrelatedtoESGanditspillarsareasprovidedbyRefinitivand
obtainedfromThompsonReuters’DataStream
Using this data, we then construct ten portfolios based on the ESG scores at the end of
June for each year in the period 2003-2019. The descriptive statistics of the ten resulting
portfolios are presented in Table 2 on the next page where #1 refers to the portfolio based on
the highest decile of ESG scores and #10 to the portfolio based on the lowest decile of ESG
scores.
13
Table 2: Descriptive Portfolio Statistics
#1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 VARIABLES !"#$&! 0.0060 0.0073 0.0061 0.0074 0.0072 0.0085 0.0092 0.0095 0.0107 0.0098 -0.0037 '"!, 0.0376 0.0378 0.0387 0.0372 0.0382 0.0433 0.0414 0.0412 0.0426 0.0397 0.0152 !#(&! 0.1201 0.1130 0.1079 0.0902 0.1081 0.1158 0.1346 0.1385 0.1358 0.1285 0.0381 !)$&! -0.1664 -0.1453 -0.1537 -0.1544 -0.1421 -0.1845 -0.1773 -0.1818 -0.1735 -0.1653 -0.0468 MeanMarketCap
24.085 17.245 15.199 15.152 13.083 10.155 7.195 5.850 4.986 4.189 14.137
MeanESGCombinedScore
74.560 62.102 54.152 47.538 41.357 36.013 30.655 25.343 19.521 11.077
42.818
WhereMean)!representstheaveragemonthlyportfolioreturnwherethemonthlyportfolioreturniscalculatedasthesumofallweightedreturnsofthestockscomprisingtheportfolio.;"! representstheportfoliostandarddeviation.<=>)!and<?@)!representthemaximumandminimumvaluesofmonthlyportfolioreturnsrespectively.WhereasMeanMarketCapreferstotheaveragemarketcapitalisationinmillionsofUSdollarsofthefirmscomprisingtheportfolio.TheMeanESGCombinedscoreistheaverageESGCombinedScoreofthestocksincludedin
theportfolio.
14
On first inspection, it seems that stocks in the lower ESG scoring portfolios have a
higher average return than those in the higher ranked portfolios. Another interesting observation
is that, based on the table, there is a relation between market capitalisation and the ESG
Combined score, where the firms included in the highest scoring portfolios have on average a
much larger market capitalisation than those in the lower end, with the difference between
portfolio number 1 and 10 being almost a factor six.
The data for the factors is obtained from Kenneth R. French’s webpage where we use
the global factors since our portfolios are not limited to specific regions and contain stocks from
all over the world. For the risk free rate we use the one month US Treasury bill rate.
To see if these observations are actually meaningful and significant, we will present the
results of our regressions in the next section.
5. Results To obtain the different Beta’s we will use Ordinary Least Squares (OLS) estimations.
In order to prevent the possible presence of heteroskedasticity affecting the results we use
White's heteroskedasticity consistent standard errors.
To test our hypothesis that ESG scores do affect returns we run the regressions for the
different models (CAPM, Fama French 3 factors, Carhart 4 factors and the Fama French Five
factor model). The results thereof can be found in the subsequent tables where #11 relates to
the portfolio with a long position in portfolio #1 and a short position in #10.
First, table 3 presents the results for the CAPM. What immediately stands out is that
both of the “extreme” portfolios yield significant alphas where interestingly, the highest ESG
scoring portfolio yields a negative alpha of 0.488%, whereas the worst portfolio in terms of
ESG does yield a significantly positive alpha of 0.2825%. This implies that a portfolio
consisting of the 10% worst ESG scoring stocks outperform the market by 0.2825 percent point.
When adding the SMB and HML factors the results from the CAPM seem to hold.
Again, the number 1 portfolio has a significantly negative alpha, albeit a lot lower at -0.106%.
As under CAPM, the worst ESG portfolio has an alpha of 0.299% at the 90% significance level.
This in turn means that the difference portfolio does yield a significantly lower alpha of -0.355%
compared of the individual portfolios of which it is comprised.
15
Table 3: CAPM Results July 2003 – June 2020
#1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 VARIABLES
!! 0.182*** 0.829*** 0.835*** 0.859*** 0.825*** 0.861*** 0.942*** 0.915*** 0.910*** 0.908*** -0.727*** (0.0688) (0.0315) (0.0279) (0.0290) (0.0254) (0.0230) (0.0375) (0.0313) (0.0324) (0.0343) (0.0875)
# -0.448** -0.129 -0.00831 -0.145 0.0140 -0.0312 0.0324 0.129 0.158 0.282** -0.729** (0.206) (0.104) (0.101) (0.102) (0.0978) (0.0889) (0.130) (0.112) (0.112) (0.138) (0.282) Observations 204 204 204 204 204 204 204 204 204 204 204 R-squared 0.080 0.864 0.868 0.876 0.872 0.902 0.844 0.868 0.866 0.808 0.429
Results are obtained using an OLS time-series regression on the CAPM. Where !! refers to the coefficient for the market risk premium. # represents the outperformance of the portfolio (in %).
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
16
Table 4: Fama French Three-Factor Results July 2003 – June 2020
#1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 VARIABLES
!! 0.00259 0.857*** 0.851*** 0.886*** 0.856*** 0.864*** 0.956*** 0.932*** 0.917*** 0.924*** -0.922*** (0.00185) (0.0311) (0.0298) (0.0325) (0.0287) (0.0237) (0.0394) (0.0324) (0.0334) (0.0361) (0.0365)
!"#$ 0.00238 -0.130*** -0.0873* -0.110** -0.143*** -0.0338 -0.0571 -0.0142 0.0289 0.0579 -0.0556 (0.00388) (0.0466) (0.0475) (0.0431) (0.0450) (0.0415) (0.0542) (0.0452) (0.0481) (0.0607) (0.0611)
!%#& 0.998*** 0.0374 0.0373 0.00797 0.0369 0.0345 0.00859 -0.0706 -0.0805* -0.174*** 1.172*** (0.00294) (0.0469) (0.0473) (0.0522) (0.0401) (0.0476) (0.0625) (0.0498) (0.0460) (0.0508) (0.0511)
# -0.106*** -0.132 -0.00624 -0.156 0.00911 -0.0235 0.0283 0.103 0.134 0.229* -0.335** (0.00925) (0.104) (0.102) (0.104) (0.0958) (0.0893) (0.129) (0.111) (0.110) (0.135) (0.137) Observations 204 204 204 204 204 204 204 204 204 204 204 R-squared 0.998 0.869 0.871 0.879 0.879 0.902 0.844 0.870 0.868 0.818 0.845
Results are obtained using an OLS time-series regression on the Fama and French (1993) Three Factor Model. Where !! refers to the coefficient for the market risk premium, !"#$ is the parameter for the return long in small stocks and short in large cap stocks. !%#& is the coefficient corresponding to the return of a portfolio long in high book-to-market value and short in low book-to-market value shares. # represents the
outperformance of the portfolio (in %) Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
17
Next we further expand our analysis by adding the momentum factor as suggested by
Carhart (1997). The results of this can be found in table 5 on the next page. The additional
momentum factor does not appear to be significant for many of the portfolios. The results with
regards to the alpha seem to comply with those obtained from the other models. Again, it seems
that the portfolio scoring highest on ESG is the only portfolio with a significant alpha whereas
the worst portfolio in terms of ESG performance is the only that yields a significantly positive
alpha.
Finally, we test the Fama French 5 factor model by including the Robust Minus Weak
and the Conservative Minus Aggressive factors. As with the previously added factors, the
corresponding betas mostly turn out to be statistically not different from zero. However, in
terms of alpha we again observe that the only significant alphas are yielded by the best and
worst ESG scoring portfolio, where the highest ESG portfolio does not meet the market
benchmark while the lowest scoring portfolio again has a positive alpha of 0.296%. This implies
that the portfolio constructed out of the percentile of lowest ESG scoring stocks does
outperform the market by almost 0.3 percent point. Subsequently, the portfolio consisting of
the top ESG scoring decile does underperform, where the difference with the market return is -
0.106 percent.
The results of the regressions seem to a relationship between ESG performance and
returns. Surprisingly however, it seems that that the best performing portfolio in terms of ESG
is also the only portfolio that has a negative alpha.
The negative alpha for the high ESG scoring portfolio could be an indication that the
stocks that comprise it are overpriced and hence do underperform in terms of returns. This could
be explained by investors preferring high ESG scoring stocks where the resulting increased
demand pushes prices past their natural levels. Contrarily, the positive alpha for the worst ESG
scoring portfolio might reflect investor’s reluctance to invest in bad ESG stocks which drives
down prices, resulting in under-priced assets. This is in line with the results of Hong and
Kacperczyk (2009), even when using the Fama French five factor model. An important note to
this is that the worst ESG scoring decile portfolio is not necessarily comprised of only sin
stocks, since these are defined as stocks belonging to specific industries whereas in our
portfolios, the selection is made not on industry but overall ESG scores. The implications are
however similar, the under-pricing of sin/low ESG assets can possibly be attributed to a lack of
(institutional) investors interest.
18
Table 5:
Carhart Four-Factor Results July 2003 - June 2020
#1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11
VARIABLES
!! 0.00200 0.847*** 0.836*** 0.875*** 0.848*** 0.859*** 0.949*** 0.919*** 0.898*** 0.904*** -0.902***
(0.00195) (0.0329) (0.0325) (0.0333) (0.0300) (0.0257) (0.0423) (0.0344) (0.0328) (0.0377) (0.0381)
!"#$ 0.00253 -0.128*** -0.0834* -0.107** -0.141*** -0.0327 -0.0554 -0.0111 0.0338 0.0631 -0.0606
(0.00386) (0.0457) (0.0458) (0.0418) (0.0446) (0.0412) (0.0535) (0.0439) (0.0450) (0.0573) (0.0580)
!%#& 0.997*** 0.0168 0.00621 -0.0144 0.0206 0.0259 -0.00454 -0.0951* -0.119** -0.215*** 1.212***
(0.00308) (0.0477) (0.0477) (0.0555) (0.0389) (0.0525) (0.0611) (0.0511) (0.0483) (0.0511) (0.0517)
!#'# -0.00209 -0.0363 -0.0546* -0.0394 -0.0286 -0.0151 -0.0231 -0.0431 -0.0676** -0.0720* 0.0699*
(0.00165) (0.0401) (0.0278) (0.0280) (0.0230) (0.0235) (0.0369) (0.0413) (0.0323) (0.0388) (0.0390)
# -0.105*** -0.125 0.00554 -0.147 0.0153 -0.0202 0.0333 0.112 0.148 0.245* -0.350**
(0.00925) (0.105) (0.103) (0.105) (0.0969) (0.0899) (0.131) (0.113) (0.110) (0.136) (0.137)
Observations 204 204 204 204 204 204 204 204 204 204 204
R-squared 0.998 0.871 0.874 0.881 0.880 0.903 0.845 0.871 0.872 0.822 0.848
Results are obtained using an OLS time-series regression on the Carhart (1997) Four Factor Model. Where !! refers to the coefficient for the
market risk premium, !"#$ is the parameter for the return long in small stocks and short in large cap stocks. !%#& is the coefficient
corresponding to the return of a portfolio long in high book-to-market value and short in low book-to-market value shares. !#'# is the
coefficient related to the momentum factor while # represents the outperformance of the portfolio (in %).
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
19
Table 6:
Fama French Five-Factor Model Results July 2003 – June 2020
#1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11
VARIABLES
!! 0.00312 0.857*** 0.837*** 0.875*** 0.848*** 0.861*** 0.932*** 0.923*** 0.900*** 0.887*** -0.884***
(0.00202) (0.0303) (0.0298) (0.0327) (0.0272) (0.0243) (0.0362) (0.0325) (0.0332) (0.0339) (0.0342)
!"#$ 0.00251 -0.128** -0.0947* -0.123*** -0.138*** -0.0333 -0.0734 -0.00516 0.0337 0.0463 -0.0438
(0.00394) (0.0519) (0.0501) (0.0464) (0.0463) (0.0441) (0.0590) (0.0495) (0.0512) (0.0598) (0.0602)
!%#& 0.996*** 0.0393 0.0737 0.0269 0.0683 0.0442 0.0647 -0.0287 -0.0186 -0.0677 1.064***
(0.00343) (0.0569) (0.0521) (0.0615) (0.0524) (0.0557) (0.0786) (0.0574) (0.0553) (0.0553) (0.0554)
!(#) 0.00132 0.0149 -0.0560 -0.0836 0.0184 -0.000364 -0.115 0.0384 0.00614 -0.105 0.106
(0.00473) (0.0793) (0.0682) (0.0737) (0.0700) (0.0608) (0.0944) (0.0814) (0.0755) (0.0905) (0.0908)
!*#+ 0.00536 -0.00875 -0.122 -0.0557 -0.115 -0.0345 -0.184 -0.155* -0.221** -0.364*** 0.369***
(0.00639) (0.0914) (0.0760) (0.0837) (0.0940) (0.0751) (0.118) (0.0880) (0.0956) (0.0971) (0.0981)
# -0.107*** -0.136 0.0228 -0.124 0.0141 -0.0201 0.0817 0.105 0.153 0.296** -0.403***
(0.00947) (0.0986) (0.101) (0.107) (0.0918) (0.0916) (0.124) (0.109) (0.107) (0.133) (0.135)
Observations 204 204 204 204 204 204 204 204 204 204 204
R-squared 0.998 0.869 0.873 0.881 0.880 0.902 0.848 0.872 0.873 0.830 0.855
Results are obtained using an OLS time-series regression on the Fama and French (2015) Five Factor Model. Where !! refers to the coefficient
for the market risk premium, !"#$ is the parameter for the return long in small stocks and short in large cap stocks. !%#& is the coefficient
corresponding to the return of a portfolio long in high book-to-market value and short in low book-to-market values. !(#) represents the
parameter of the return of a portfolio long in robust profit shares and short in weak profit shares. !*#+ is the expression of the coefficient of the
return of a portfolio long in conservative investment portfolio and short in aggressive investment portfolio. While # represents the
outperformance of the portfolio (in %).
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
20
6. Robustness
To test the robustness of our results we will repeat the regressions of the Fama French
Five Factor model but this time using equally weighted portfolios, in order to test whether or
not the results might be biased due to the increased influence of larger firms in value weighted
portfolios. The results of these regressions can be found in Appendix B
The results for the equally weighted portfolios seem to mostly match those of the value
weighted portfolio, the main difference being that for the #10 portfolio, there is no significant
alpha, and hence no advantage in investing in low ESG scoring stocks. For the best scoring
portfolio in terms of ESG, there still is a significant alpha of -0.107%. This is in part in line
with the hypothesis that ESG does have an effect on portfolio returns.
7. Conclusion Given the results obtained from our estimation there seems to be evidence for some sort
of relationship between ESG and portfolio returns. Both the highest and lowest ESG scoring
portfolios and subsequently the portfolio long in the high ESG and short in the low ESG
portfolios generate significant alphas across the different models used for our estimations. For
the remaining portfolios however, there seems to be no outperformance in any direction.
Starting with the high ESG portfolio, using the CAPM the alpha is estimated at -0.45%
which is a lot lower than the alpha of -0.11% that results from the other multi-factor model
estimations. The fact remains that this negative alpha is sustained throughout the different
estimation models implying that there is a significant negative effect of ESG on portfolio
returns.
For the lowest ESG portfolio however, thing are quite different. Starting at the CAPM
where alpha is estimated at 0.28%. This number slightly decreases (but remains significant)
when estimating the Fama and French three-factor model and the Carhart four-factor model and
even increases to a value of 0.30% when estimated using the Fama and French five-factor
model. This seems to be in line with previous literature regarding sin anomaly, where sin stocks
do outperform the market. However, this result could not be replicated in the robustness check
which used equally weighted portfolios. Also, as noted earlier, low ESG does not necessarily
equate to sin stocks given its definition of being limited to specific industries.
21
Overall, we can conclude that ESG does have an effect on portfolio returns. One should
however take caution with this conclusion since the effect appears to only be significant at the
extremes of ESG performance. Moreover, the results are limited and provide a basis for further
research, the details of which will be discussed in the next section.
8. Discussion The results of this paper should be carefully interpreted and invite for further research
into this topic. In our setup we made no distinction between regions and opted for a global
approach. It would be interesting to see if there are differences between regions in terms of the
relation between ESG and financial performance. This could for example be due to differences
in local preferences or focus on non-financial characteristics of portfolios.
Another option for further research is into the difference between industries previous
studies often focussed specifically on vice industries and found that these tend to give excess
returns, ignoring any of the ESG characteristics of the firms active in these industries. By taking
a broader approach looking beyond industry but fo
Another area to further adres would be the effect of ESG on returns during times of
crisis. Do the more long-term oriented high ESG shares outperform low ESG shares during
economic downturn as suggested by some studies, and do these results hold when using the
relatively new Fama and French (2015) five-factor model. Or are sin stock which tend to be
often associated with fast moving consumer goods less dependent on economic cycles?
Finally, in this paper we used the ESG Combined Score as a criterium in constructing
our portfolio, it would be interesting to see if results would differ for the different pillars of
ESG, can alpha be achieved by selecting stocks on the basis of their performance in on of the
ESG pillars instead of focusing on overall ESG performance.
In conclusion ESG does affect portfolio returns, but more questions remain unanswered
and invite further research into this area.
22
9. Apendix
Apendix A: Number of Firms
WhereTotalreferstothetotalnumberoffirmsavailableforeachgivenyear.#1upuntil#11refertothenumberoffirmsincludedineachrespectiveportfolioforeachgivenyear.
Year
Total #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11
2003 867 86 86 86 87 87 87 87 87 87 87 173
2004 1,583 159 159 159 158 158 158 158 158 158 158 317
2005 1,988 198 199 199 199 199 199 199 199 199 198 396
2006 2,009 201 200 201 201 201 201 201 201 201 201 402
2007 2,127 213 213 212 212 212 213 213 213 213 213 426
2008 2,530 253 253 253 253 253 253 253 253 253 253 506
2009 2,971 297 297 297 297 297 297 297 297 297 298 595
2010 3,459 346 346 346 346 346 345 346 346 346 346 692
2011 3,501 350 350 350 350 350 351 350 350 350 350 700
2012 3,574 358 358 358 358 357 357 357 357 357 357 715
2013 3,710 371 371 371 371 371 371 371 371 371 371 742
2014 3,823 383 383 382 382 383 382 382 382 382 382 765
2015 4,556 455 455 455 455 456 456 456 456 456 456 911
2016 5,286 529 529 529 529 529 529 528 528 528 528 1,055
2017 5,645 564 564 565 565 564 564 565 565 565 564 1,128
2018 6,258 626 626 626 626 626 626 626 625 625 626 1,252
2019 6,206 620 620 620 620 621 621 621 621 621 621 1,241
23
Apendix B: Fama French Five-Factor Model Results July 2003 – June 2020 Equally Weighted Portfolios
#1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 VARIABLES
"! 0.00312 0.889*** 0.876*** 0.900*** 0.888*** 0.904*** 0.883*** 0.906*** 0.876*** 0.871*** -0.868*** (0.00202) (0.0383) (0.0387) (0.0439) (0.0415) (0.0422) (0.0417) (0.0422) (0.0377) (0.0402) (0.0403)
""#$ 0.00251 0.0280 0.0814 0.0610 0.138** 0.174*** 0.228*** 0.240*** 0.278*** 0.253*** -0.250*** (0.00394) (0.0663) (0.0681) (0.0624) (0.0647) (0.0600) (0.0566) (0.0533) (0.0551) (0.0574) (0.0575)
"%#& 0.996*** 0.141* 0.176** 0.140 0.135 0.146* 0.156* 0.133 0.0931 0.154** 0.843*** (0.00343) (0.0792) (0.0767) (0.0851) (0.0848) (0.0860) (0.0822) (0.0844) (0.0787) (0.0772) (0.0768)
"'#( 0.00132 0.0109 -0.0387 -0.0530 -0.0120 -0.0473 -0.0748 -0.0412 0.000914 -0.0511 0.0524 (0.00473) (0.0938) (0.0946) (0.0879) (0.0882) (0.0846) (0.0838) (0.0758) (0.0786) (0.0799) (0.0797)
")#* 0.00536 -0.137 -0.184* -0.150 -0.216** -0.195* -0.221** -0.167* -0.191* -0.291*** 0.296*** (0.00639) (0.109) (0.110) (0.113) (0.108) (0.102) (0.101) (0.0994) (0.107) (0.0950) (0.0959)
# -0.107*** -0.150 -0.00206 -0.0483 -0.0378 0.0425 0.0833 0.121 0.0773 0.147 -0.254* (0.00947) (0.126) (0.128) (0.128) (0.123) (0.121) (0.123) (0.121) (0.118) (0.132) (0.133) Observations 204 204 204 204 204 204 204 204 204 204 204 R-squared 0.998 0.849 0.849 0.859 0.865 0.882 0.880 0.893 0.884 0.871 0.865
Results are obtained using an OLS time-series regression on the Fama and French (2015) Five Factor Model. Where "! refers to the coefficient for the market risk premium, ""#$ is the parameter for the return long in small stocks and short in large cap stocks. "%#& is the coefficient corresponding to the return of a portfolio long in high book-to-market value and short in low book-to-market values. "'#( represents the
parameter of the return of a portfolio long in robust profit shares and short in weak profit shares. ")#* is the expression of the coefficient of the return of a portfolio long in conservative investment portfolio and short in aggressive investment portfolio. While # represents the
outperformance of the portfolio (in %). Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
24
10. References
- Auer, B. R., & Schuhmacher, F. (2016). Do socially (ir) responsible investments pay? New evidence from international ESG data. The Quarterly Review of Economics and Finance, 59, 51-62.
- Broadstock, D. C., Chan, K., Cheng, L. T., & Wang, X. (2020). The role of ESG performance during times of financial crisis: Evidence from COVID-19 in china. Finance research letters, 101716.
- Carhart, C. (1997). On persistence in mutual fund performance, Journal of Finance, 45(5), 57-
82.
- Chawana, M. (2014). Socially responsible investing returns: Evidence from South Africa, 2004-2012. Journal of Economic and Financial Sciences, 7(1), 103-126.
- Dorfleitner, G., Utz, S., & Wimmer, M. (2013, October). Where and when does it pay to be good? A global long-term analysis of ESG investing. In 26th Australasian Finance and Banking Conference.
- Fama, E.F., French, K.R. (1993). Common risk factors in the returns on stocks and bonds,
Journal of Financial Economics, 33(1), 3-56.
- Fama, E. F., & French, K. R. (2015). A five-factor asset pricing model. Journal of Financial
Economics, 116(1), 1-22.
- Friede, G., Busch, T., & Bassen, A. (2015). ESG and financial performance: aggregated evidence from more than 2000 empirical studies. Journal of Sustainable Finance & Investment, 5(4), 210-233.
- Global Sustianable Investment Alliance (2018) Global Sustainable Investment Review 2018,
Retrieved from http://www.gsi-alliance.org/wp-
content/uploads/2019/03/GSIR_Review2018.3.28.pdf [Accessed October 13, 2020]
- Hoepner, A. G. F. and Zeume, S. (2014) ‘Fiduciary Duty and ‘Sin Stocks’: is vice really
nice?’ In J. P. Hawley, A. G. F. Hoepner, K. L. Johnson, J. Sandberg, and E. J. Waitzer (eds.)
Handbook of institutional investment and fiduciary duty:181-206 .Cambridge: Cambridge
University Press.
- Hong H., and Kacperczyk M. (2009). The Price of Sin: the Effects of Social Norms on
Markets, Journal of Financial Economics, 93(1), 15-36.
- Humphrey, J. E., Lee, D. D., & Shen, Y. (2012). Does it cost to be sustainable?. Journal of Corporate Finance, 18(3), 626-639.
- Jain, M., Sharma, G. D., & Srivastava, M. (2019). Can sustainable investment yield better financial returns: A comparative study of ESG indices and MSCI indices. Risks, 7(1), 15.
- Mǎnescu, C. (2011). Stock returns in relation to environmental, social and governance performance: Mispricing or compensation for risk?. Sustainable development, 19(2), 95-118.
- Merton, R.C. (1987). A Simple Model of Capital Market Equilibrium with Incomplete
Information. The Journal of Finance, 42(3), 483-510.
- Nauman, B. (2020, September 14) ESG surges as investors search for better corporate citizens. Financial Times, Retrieved from: https://www.ft.com/content/20f6c929-2fbf-47d5-973c-8c18607fc604 [Accessed October 10, 2020]
25
- Nofsinger, J., & Varma, A. (2014). Socially responsible funds and market crises. Journal of Banking & Finance, 48, 180-193.
- Singh, A. (2020). COVID-19 and safer investment bets. Finance research letters, 101729. - Torre, M. L., Mango, F., Cafaro, A., & Leo, S. (2020). Does the ESG Index Affect Stock
Return? Evidence from the Eurostoxx 50. Sustainability, 12(16), 6387. - Richey, G. (2017), "Fewer reasons to sin: a five-factor investigation of vice stock
returns", Managerial Finance, Vol. 43 No. 9, pp. 1016-1033.
- Sahut, J.-M., Pasquini-Descomps Hélène, Cohendet, P., & Mazouz, B. (2015). Esg impact on market performance of firms: international evidence. Management International, 19(2), 40–63.
- Salaber, J. (2013). Religion and returns in Europe, European Journal of Political Economy,
32(1), 149-160.