Journal of Asset Management (2020) 21:52–69

https://doi.org/10.1057/s41260-019-00139-z

ORIGINAL ARTICLE

Integrating sustainability risks in asset management: the role of ESG

exposures and ESG ratings
Benjamin Hübel1 · Hendrik Scholz1

Revised: 16 August 2019 / Published online: 21 November 2019

© Springer Nature Limited 2019

Abstract
The rising sustainability awareness among regulators, consumers and investors results in major sustainability risks of firms.
We construct three ESG risk factors (Environmental, Social and Governance) to quantify the ESG risk exposures of firms.
Taking these factors into account significantly enhances the explanatory power of standard asset pricing models. We find
that portfolios with pronounced ESG risk exposures exhibit substantially higher risks, but investors can compose portfolios
with lower ESG risks while keeping risk-adjusted performance virtually unchanged. Moreover, investors can measure the
ESG risk exposures of all firms in their portfolios using only stock returns, so that even stocks without qualitative ESG
information can be easily considered in the management of ESG risks. Indeed, strategically managing ESG risks may result
in potential benefits for investors.

Keywords ESG ratings · ESG exposures · Risk management · Stock returns · CSR

JEL Classification G11 · G12 · G14 · G23

Introduction Promoting sustainability by achieving these SDGs

requires profound changes in the economic environment.
In 2017, insurers paid a record of $134bn in compensation Consumers, investors and other stakeholders expect firms
for natural disasters (Aon Benfield 2017). Highlighting the to adopt sustainable business models. Also, regulatory
importance of effective governance, Volkswagen’s emissions authorities are driving the integration of ESG considera-
scandal in 2015 resulted in equity market value losses of tions by directing private capital flows toward sustainable
more than $20bn for Volkswagen within five trading days investments.1 For instance, in 2019, the European Parliament
and substantial losses for Volkswagen’s competitors due is expected to adopt a fundamental directive intended to put
to spillover effects (Barth et al. 2019a). Also, population “ESG considerations at the heart of the financial system”
growth, globalization and inequality impose severe sus- (Directive 2016/2341). Risks arising from these effects are
tainability challenges. Recognizing the importance of such hardly diversifiable, as they affect the whole economic envi-
environmental, social and governance (ESG) considerations, ronment, not just single firms or industries (Scholtens 2017).
the United Nations formalized the Sustainable Development To be effective, risk management systems need to con-
Goals (SDGs, United Nations 2015) which were adopted by sider ESG risks when assessing portfolios. For investors,
governments around the world. risk management is a main motivation to consider ESG in
their investment decisions. However, a lack of quantitative
ESG data which is comparable, easily accessible and of high
quality slows the integration of ESG (CFA Institute 2017).
* Benjamin Hübel
[email protected] We therefore propose to apply return-based ESG expo-
sures to measure ESG risks and integrate these into asset
Hendrik Scholz
[email protected] management. In doing so, we contribute to the existing

1 1
Chair of Finance and Banking, Friedrich-Alexander- Private investments are needed to close the €180-billion gap in
Universität Erlangen-Nürnberg (FAU), Lange Gasse 20, additional yearly investments needed to meet the SDGs (European
90403 Nürnberg, Germany Commission 2018).

Vol:.(1234567890)
Integrating sustainability risks in asset management: the role of ESG exposures and ESG ratings 53

literature in three ways. First, we explore three ESG-related 2016; El Ghoul and Karoui 2017). Sidestepping the issue,
risk factors in order to capture ESG risk exposures of firms. authors often assume that rated stocks appropriately rep-
Second, we derive conclusions regarding the link between resent unrated stocks (Wimmer 2013; Borgers et al. 2015;
ESG and performance by composing portfolios based on Ameer 2016; Auer 2016). Neglecting unrated stocks reduces
ESG ratings and ESG exposures.2 Third, we investigate the investment universe which could impact performance
whether the performance of stocks without ESG ratings through a lack of diversification. Also, ESG rating agencies
affects these conclusions. predominantly cover large firms resulting in a systematic
Our empirical study applies an augmented Fama and lack of small firms in ESG portfolios. Unlike ESG ratings,
French (2015) five-factor model to estimate ESG risk expo- stock returns are broadly available enabling us to calculate
sures of firms. For each ESG pillar, we develop one ESG- ESG risk exposures for all firms. Therefore, composing ESG
related zero-investment factor that covers the return differ- portfolios based on ESG exposures enables us to consider
ences between firms with high ESG ratings (long) and firms both rated and unrated stocks.
with low ESG ratings (short). In the same spirit, Statman and Studying a comprehensive sample of European stocks
Glushkov (2016) develop two ESG-related factors to clas- from 2003 to 2016, we derive four main results. First, ESG
sify mutual funds. In contrast to our study, their factors are ratings and ESG exposures of firms correlate positively but
related to ESG-motivated exclusions of sin industries and do not perfectly. Differences between ratings and exposures
not investigate the three ESG pillars separately.3 Especially, may be caused by the industry benchmarking of ESG rat-
Görgen et al. (2018) develop a carbon factor to measure the ings and by ESG risks which are not assigned to firms by
sensitivities of firms to stock market’s time-varying percep- ESG rating agencies. Second, the environmental risk fac-
tion of risks arising from the transition toward a carbon-free tor yields a significant positive annual mean return of three
economy. This carbon factor covers risks of firms regarding percent, indicating that “brown” firms outperform “green”
carbon emissions along the value chain, public perception of firms. Third, the social risk factor shows significant and neg-
carbon policy and adaptability of low-carbon business mod- ative returns during recessions. This suggests that the least
els. In contrast, our environmental factor is more generally social firms underperform the most social firms, which is
defined, relying on the overall environmental performance. in line with a “flight to quality” effect during crisis periods.
In addition to carbon emissions, this includes resource use Fourth, ESG factors significantly enhance the explanatory
(for example, energy consumption and waste treatment), power of standard asset pricing models. However, our results
water emissions as well as firms’ capabilities to foster inno- do not point to a systematic ESG-related risk premium or
vations reducing overall environmental costs. Henriksson discount. While very pronounced ESG exposures are associ-
et al. (2019) construct an ESG-related factor to overcome the ated with higher portfolio risks, we show that investors can
sparse voluntary ESG data reported by firms in the S&P 500. still reduce ESG risk exposures of their portfolios without
Unlike our factors, they do not distinguish between the three harming performance.
ESG pillars and they do not rely on ESG ratings but on mate- The remainder of the paper is organized as follows: “ESG
rial ESG information published by the firms themselves. exposures versus ESG ratings” section describes and com-
In general, ESG ratings are usually available for a lim- pares the concepts of ESG ratings and ESG exposures. “ESG
ited number of stocks only. In the empirical literature on factors and stock returns” section presents the data and intro-
equity funds, usually 20–40 percent of equity portfolios duces our three ESG risk factors. Following a descriptive
remain unrated (Kempf and Osthoff 2008; Auer 2016; Henke analysis of the factors, we carry out asset pricing tests for
ESG decile portfolios. “Portfolio performance under consid-
eration of ESG” section delivers risk and return properties
2
of ESG portfolios based on ESG ratings and ESG exposures.
There is an ongoing debate regarding the impact of ESG integra-
In addition, we examine if unrated stocks affect our results.
tion on financial performance. Though there is evidence for CSR
affecting corporate performance [for example, Hong and Kacper- Subsequently, we investigate ESG risks from an investor’s
czyk (2009); El Ghoul et al. (2011); Chava (2014) and Eccles et al. perspective. “Conclusion” section concludes.
(2014)], literature investigating the performance of ESG portfolios
shows mixed results. Papers covering the US market [for example,
Derwall et al. (2005); Galema et al. (2008); Dorfleitner et al. (2018);
Gloßner (2018) and Amiraslani et al. (2019)] and the European mar- ESG exposures versus ESG ratings
ket (Brammer et al. 2006; Mollet and Ziegler 2014; Auer 2016; Auer
and Schuhmacher 2016 and Barth et al. 2019b) indicate a partly posi- ESG concerns such as natural resources, supply chain
tive relation between CSR and performance for the US, whereas the
flow and labor reliability, as well as evolving governance
evidence for Europe does not show a significant performance impact.
3 regulations, play an important role when assessing invest-
Similarly, Jin (2018) applies one aggregated ESG-related factor to
evaluate U.S. equity funds and finds that fund managers tend to hedge ment risks. Hence, it is of crucial importance for investors
ESG-related systematic risks. to consider ESG risks in their risk management systems.
54 B. Hübel, H. Scholz

Accordingly, access to consistent, comparable and reliable potential greenwashing and incomplete data (Parguel et al.
ESG information has become a precondition to making 2011).5
investment decisions. In addition, commonly used ESG rating agencies, such
To assess a firm’s level of sustainability, researchers and as MSCI, Sustainalytics or Thomson Reuters, provide
practitioners usually rely on ESG ratings. Based on publicly relative ESG ratings using industry adjustments. ESG rat-
available information, private rating agencies aggregate a ings provide insights regarding the ESG performance of a
large number of ESG-related indicators to derive numeric firm relative to its industry peers, whereas ESG exposures
ESG labels for each firm. More specifically, rating agen- deliver absolute exposures to ESG risks. For instance,
cies usually derive ratings for each of the three ESG pillars consider the oil and gas firm Royal Dutch Shell. Numer-
separately, as well as collating one aggregated ESG rating. ous sustainability activities make Shell a leader within its
In this paper, we use each of the three ESG ratings separately industry resulting in high environmental ratings. Accord-
to determine the ESG risk exposures of firms based on cor- ing to its environmental ratings, Shell is among the best
relations between stock returns and the respective ESG risk stocks in our sample. Nevertheless, Shell’s value chain is
factors. Importantly, these ESG exposures are not designed associated with extensive greenhouse gas emissions and
to replace ESG ratings. Rather, they reflect firms’ sensitivi- considerable use of materials, energy and water. Shell’s
ties to ESG risks triggered by the rising sustainability con- stock returns are therefore associated with high environ-
sideration on capital markets. mental risk exposures. Accordingly, Shell is ranked low
To get a sense of the relationship between ESG ratings in our sample, with regard to its (absolute) environmental
and the ESG exposures of firms, we calculate Spearman rank exposure.
correlations of the two over time and for each pillar. These In summary, investors usually apply ESG ratings to assess
correlations range around 0.25. Although the correlations ESG risks of firms. These ESG ratings provide numeric,
are significantly positive, there are still notable differences industry-benchmarked ESG labels. We propose to enhance
between rankings based on ESG ratings and ESG expo- investors’ ESG risk assessment using ESG exposures based
sures.4 These differences might arise when publicly avail- on standard asset pricing techniques that reflect the valuable
able information (ESG reports, media coverage, etc.) implies informational content of stock returns.
a high level of ESG, whereas information reflected in stock
returns shows a positive risk exposure to ESG factors or vice
versa. We identify two main causes of such discrepancies: ESG factors and stock returns
(1) unassigned ESG risks and/or (2) industry adjustments
of ESG ratings. Data
Unassigned ESG risks occur if ESG ratings do not
reflect firm-specific characteristics causing a (positive or As a measure of corporate sustainability, we use ESG
negative) correlation of the firm’s returns and the ESG ratings from the Thomson Reuters ESG database which
factors’ returns. For instance, consider a grocery retailer is a replacement and enhancement of the ESG database
that is highly responsible based on common ESG measures ASSET4 (Utz and Wimmer 2014; Chatterji et al. 2015;
such as eco-efficient supply chain management, safe and Dorfleitner et al. 2015). Thomson Reuters provides trans-
satisfied workforce and commitment to best practice gov- parent, accurate and comparable ESG ratings for a global
ernance principles. Based on comprehensive ESG reports, universe of more than 4500 firms (Stellner et al. 2015).
the firm might receive very high ESG ratings. However, if Based on quantitative and qualitative ESG data (ESG
the grocery branches are located on the Las Vegas Strip, reports, NGO announcements, etc.), Thomson Reuters
the firm’s returns correlate with the neighboring gambling derives more than 400 firm-level ESG measures. These
casinos. Consequently, the firm’s ESG exposure would measures are aggregated to three firm-level ESG pillar rat-
indicate high ESG risks (Statman 2016). Irrespective of ings. A percentile rank methodology leads to relative ESG
the question of whether rating agencies should punish the ratings ranging from the lowest percentile (lowest level
firm for selling products in Las Vegas, ESG exposures can of ESG) to the highest percentile (highest level of ESG).6
help investors to be aware of the ESG risks associated with
the firm. Such unassigned ESG risks can also be triggered
or strengthened by ESG information being published by
5
the target firm itself. This makes ESG ratings subject to Although we use ESG ratings in the construction of the ESG fac-
tors, ESG exposures should be largely independent of greenwashing
and incomplete data due to diversification effects.
6
E- and S-ratings for each firm are benchmarked against the firm’s
4
Kendall’s rank correlations are slightly lower in magnitude, but p industry peers. Governance scores are calculated against the firm’s
values remain below 1%. country of headquarter.
Integrating sustainability risks in asset management: the role of ESG exposures and ESG ratings 55

Table 1  Data and descriptive statistics

Year 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Panel A: # of stocks
All stocks 1019 980 981 1018 1023 1057 1034 918 946 1113 1058 1061 1070 1066
Rated 415 432 600 755 778 815 825 789 806 839 824 829 838 837
Unrated 604 548 381 263 245 242 209 129 140 274 234 232 232 229
Min 5th P 1st Q Median Mean 3rd Q 95th P Max

Panel B: descriptive statistics of ESG ratings

ENV 13.39 28.60 46.73 59.62 58.70 71.94 84.39 96.32
SOC 7.36 26.72 44.62 57.57 56.42 68.76 82.83 93.09
CGV 3.92 21.14 36.62 49.30 48.96 60.47 75.57 97.16

This table shows descriptive statistics of the sample stocks and their Thomson Reuters ESG ratings. Panel A reports the numbers of all index
constituents in the STOXX Europe Total Market index, as well as the number of rated and unrated stocks at the end of each year from 2003 to
2016. Panel B shows descriptive statistics on the time series means of environmental, social and governance ratings for all rated stocks between
2003 and 2016

Our empirical analysis focuses on the European stock Factor construction and estimation of ESG
market, as Europe accounts for over half of the assets man- exposures
aged sustainably worldwide (GSIA 2017). We use yearly
constituents of the STOXX Europe Total Market Index We examine each ESG pillar (ENV, SOC and CGV) sepa-
which covers approx. 95 percent of free float market capi- rately instead of considering them together as one aggre-
talization across 17 European countries from EIKON. gated ESG factor, as mixing the three pillars might have
Panel A of Table 1 shows the total number of stocks, confounding effects (Galema et al. 2008). For instance, a
divided into stocks with and without ESG ratings at each firm with well-developed governance structures and weak
year end from 2003 to 2016. The total number of stocks employee relations may show an average ESG rating
ranges between 918 and 1113. The number of rated stocks although neither its G- nor its S-rating is close to average. In
increases from 415 in 2003 to 837 in 2016. The beginning addition, the firm’s stock returns may face offsetting impacts
of the sample period is determined by the availability of making it very difficult to assess the firm’s ESG risk expo-
the Thomson Reuters ESG ratings. Importantly, Thomson sure based on its returns—even though it is clear that the
Reuters did not change the rating methodology during firm is exposed to social risks. For these reasons, ESG pillars
the sample period. Panel B of Table 1 shows descriptive may partly overlap under certain circumstances leading to a
statistics regarding the ESG ratings. trade-off between specificity and cleanness.
For all stocks, we downloaded daily and monthly The ENV factor represents the returns of a zero-invest-
total returns, closing prices, market values (all in Euro), ment portfolio with long positions in firms with low envi-
dividend yields and Industry Classification Benchmark ronmental ratings and short positions in firms with high
codes from EIKON. Concerning stock returns, we apply environmental ratings. More precisely, like Görgen et al.
usual corrections following Ince and Porter (2006). The (2018) introducing a carbon factor, we follow the Fama and
returns of the five Fama and French factors, as well as French (1993) methodology when constructing our ESG fac-
the momentum factor, are downloaded from Kenneth tors. To determine the ENV factor, we unconditionally sort
R. French data library. 7 As risk-free rates, we use the stocks into six portfolios based on their market capitalization
1-month EURIBOR and the daily EONIA from EIKON. and their environmental rating. As breakpoints, we apply
the median size and terciles of the respective environmental
ratings. We calculate monthly value-weighted returns for
four of the six portfolios: small size and low environmental
rating (SL), big size and low environmental rating (BL),
small size and high environmental rating (SH) and big size
and high environmental rating (BH).8 As all ESG ratings are
7
We thank Kenneth R. French for supplying this data for Europe at
http://mba.tuck.dartm​outh.edu/pages​/facul​ty/ken.frenc​h/data_libra​
8
ry.html. Since these European factor returns are calculated in US dol- The number of stocks in the portfolios SL, BL, SH and BH ranges
lars, we converted them into Euro following Glück et al. (2019). between 69 in 2003 to 139 in 2016.
56 B. Hübel, H. Scholz

Fig. 1  Cumulative returns of the 60%

ESG factors. This figure shows ENV SOC CGV
the cumulative monthly returns 50%
of the ESG factors based on
monthly returns for the period 40%
2003–2016. The gray-shaded
area denotes the financial crisis
from May 2007 to February 30%
2009
20%

10%

0%

-10%

updated on a yearly basis, we also update the sorting of the To determine a firm’s return-based ESG exposures for a
portfolios on a yearly basis. Finally, we obtain the return of given year T, we perform the 9F-ESG model (Eq. 3) using
the ENV factor in month t: daily returns covering the previous year T − 1. We use the
coefficients 𝛽iENV , 𝛽iSOC , 𝛽iCGV as ESG exposures for each
(1)
( ) ( )
ENVt = 0.5 SLt + BLt − 0.5 SHt + BHt firm.
The ENV factor can be interpreted as time-varying mar-
ESG factor returns
ket valuation of ENV risks measured as the return difference
between “brown” firms and “green” firms. The SOC factor
To gain first insights regarding the ESG factors, Fig. 1 shows
and the CGV factor are constructed analogously, however,
cumulative returns of these factors from 2003 to 2016. The
based on social and governance ratings, respectively.
cumulative returns for all three factors are positive and com-
As basis for estimating the ESG exposures of firms, we
parable in their magnitude from 2003 until the beginning of
apply the Fama and French (2015) five-factor model aug-
the financial crisis in 2007.
mented by the momentum factor (Carhart 1997), hereafter
Hence, low ESG firms outperformed high ESG firms,
6F-FFC model:
regardless of the ESG pillar. During the financial crisis
eri,t = 𝛼i + 𝛽iMKT MKTt + 𝛽iSMB SMBt (Fig. 1, gray area, 05/2007–02/2009), the ENV factor experi-
enced considerable positive returns, whereas the SOC factor
+ 𝛽iHML HMLt + 𝛽iRMW RMWt
lost its previous positive returns. Similarly, the CGV factor
+ 𝛽iCMA CMAt + 𝛽iMOM MOMt + 𝜀i,t (2) shows slightly negative returns during the crisis. During the
post-crisis period from 2010 to 2016, the ENV factor con-
where eri,t is the excess return of stock i in month t, 𝛼i can be tinued to rise, whereas the SOC and CGV factors failed to
interpreted as abnormal performance of stock i and MKTt show a clear trend.
is the excess return of the market. SMBt , HMLt , RMWt , Table 2 presents descriptive statistics and Pearson cor-
CMAt and MOMt are the returns of the five Fama and French relations of the ESG factors, the five Fama and French fac-
factors and the momentum factor. 𝜀i,t is an error term with tors and the Carhart factor based on monthly returns. On
zero mean. average, the ESG factors delivered positive monthly returns,
To determine the ESG exposures for stock i, we add three indicating that low ESG firms performed better than high
ESG-related factors to the 6F-FFC model, resulting in a ESG firms during our total sample period (Panel A). The
9F-ESG model: ENV factor yielded a mean return of 26 basis points (approx.
eri,t = 𝛼i + 𝛽iMKT MKTt + 𝛽iSMB SMBt + 𝛽iHML HMLt 3.12 percent per year) which is significant at the 5% level.
The SOC and CGV factors show mean returns of seven and
+ 𝛽iRMW RMWt + 𝛽iCMA CMAt + 𝛽iMOM MOMt five basis points which are not statistically different from
+ 𝛽iENV ENVt + 𝛽iSOC SOCt + 𝛽iCGV CGVt + 𝜀i,t (3) zero. The ESG factors exhibit positive correlations, while
the variance inflation factors based on the 9F-ESG model
show a maximum of 3.99, indicating that multicollinearity
does not largely affect our results when applying the 9F-ESG
model (Eq. 3).
Integrating sustainability risks in asset management: the role of ESG exposures and ESG ratings 57

Table 2  Descriptive statistics and correlations of the Fama and French factors and the ESG factors
Mean t-stat SD Min Max VIF SMB HML RMW CMA MOM ENV SOC CGV

Panel A: total sample period (N = 168)

MKT 0.62 (1.965) 4.09 − 13.89 13.84 1.99 − 0.05 0.44 − 0.35 − 0.28 − 0.46 − 0.27 − 0.11 − 0.08
SMB 0.28 (1.965) 1.84 − 5.16 5.10 1.29 0.04 − 0.12 − 0.18 − 0.08 0.29 0.34 0.27
HML 0.10 (0.608) 2.19 − 4.43 7.33 3.99 − 0.77 0.27 − 0.46 − 0.57 − 0.39 0.02
RMW 0.32 (2.840) 1.48 − 3.90 4.31 2.89 − 0.25 0.44 0.38 0.13 − 0.04
CMA 0.15 (1.445) 1.38 − 3.01 5.78 1.59 0.16 − 0.23 − 0.25 − 0.05
MOM 0.87 (2.877) 3.91 − 26.91 12.97 1.75 0.29 0.09 0.15
ENV 0.26 (2.178) 1.54 − 6.64 4.51 3.07 0.75 0.20
SOC 0.07 (0.597) 1.42 − 6.78 3.95 2.68 0.22
CGV 0.05 (0.649) 0.95 − 2.57 2.76 1.18
Panel B: recession periods (N = 25)
MKT − 3.56 (− 7.888) 2.26 − 13.89 6.11 5.73 0.01 0.30 − 0.06 − 0.58 − 0.53 0.08 0.12 0.02
SMB − 0.32 (− 1.647) 0.98 − 5.16 5.10 2.67 − 0.13 0.01 − 0.56 − 0.10 0.62 0.49 0.25
HML − 0.45 (− 3.146) 0.71 − 4.28 2.46 2.48 − 0.65 0.05 − 0.39 − 0.29 − 0.14 0.36
RMW 0.71 (6.631) 0.54 − 1.36 4.31 2.55 0.00 0.39 0.10 − 0.13 − 0.31
CMA 0.78 (4.819) 0.81 − 2.58 5.78 4.11 0.16 − 0.46 − 0.48 − 0.09
MOM 0.71 (6.631) 0.54 − 1.36 4.31 4.32 0.09 − 0.07 0.16
ENV 0.52 (3.772) 0.69 − 1.81 4.51 3.36 0.74 0.21
SOC − 0.42 (− 4.052) 0.52 − 2.85 2.66 2.72 0.15
CGV 0.06 (0.601) 0.51 − 2.14 2.76 2.06
Panel C: expansion periods (N = 143)
MKT 1.35 (4.674) 3.46 − 10.36 13.84 1.79 − 0.17 0.47 − 0.40 − 0.09 − 0.42 − 0.36 − 0.27 − 0.12
SMB 0.38 (2.740) 1.68 − 4.25 4.82 1.23 0.06 − 0.14 0.01 − 0.05 0.21 0.30 0.29
HML 0.20 (1.063) 2.24 − 4.43 7.33 4.39 − 0.78 0.37 − 0.46 − 0.62 − 0.45 − 0.05
RMW 0.26 (2.029) 1.51 − 3.90 3.40 3.06 − 0.35 0.44 0.42 0.18 0.02
CMA 0.05 (0.441) 1.23 − 3.01 3.90 1.53 0.13 − 0.18 − 0.17 − 0.04
MOM 0.57 (1.749) 3.93 − 26.91 7.16 1.72 0.32 0.15 0.15
ENV 0.21 (1.696) 1.50 − 6.64 3.37 3.40 0.77 0.20
SOC 0.15 (1.261) 1.43 − 6.78 3.95 2.97 0.24
CGV 0.05 (0.621) 0.87 − 2.57 2.11 1.17

This table presents descriptive statistics and Pearson correlations of the Fama and French factors, the Carhart factor and the ESG factors in Panel
A. The results are calculated based on monthly returns for the period from 2003 to 2016. Mean, standard deviation, min and max are reported
in percent. Bold values indicate t-values significant at the 10% level and significant correlations larger than 0.5 in absolute values. Following
Newey and West (1987, 1994), we adjust standard errors for autocorrelation and heteroscedasticity. Panel B and C show the results from reces-
sion and expansion subsamples. We divide the total period in recessions and expansion periods. Using a peak-and-through approach, we iden-
tify March 2002 to March 2003 and May 2007 to February 2009 as recessions months. The remaining months are considered to be expansion
months

The previous literature also indicates that ESG-related C of Table 2 show descriptive statistics of the ESG factors
returns may depend on market states. High ESG firms for recession and expansion sub-periods, respectively.
may yield relative low returns in upward moving markets The ENV factor exhibits significantly positive returns in
but receive relative high returns in downward markets both market states, indicating that “brown” firms signifi-
(Nofsinger and Varma 2014; Lins et al. 2017). Following cantly outperformed “green” firms. Notably, this outper-
Nofsinger and Varma (2014), we define recessions by the formance during recessions is about twice as high as during
peaks and troughs of the STOXX Europe total return index, expansions. In line with the previous literature (for exam-
according to which two recession periods can be identified ple, Edmans 2011; Lins et al. 2017), we find that the SOC
in our sample timespan: the bursting of the dotcom bubble factor shows highly significant and negative returns during
from March 2002 to March 2003 and the global financial recessions (− 0.42 percent per month or − 5 percent p.a.),
crisis from May 2007 to February 2009. The remaining i.e., firms with high SOC ratings significantly outperform
months are considered periods of expansion. Panels B and firms with low SOC ratings. These results are consistent
58 B. Hübel, H. Scholz

with a “flight to quality” effect as reported by Dong et al. deciles. F-tests show statistical significance at the 1% level
(2019) for the US market. During crisis periods, investors for more than half of the decile portfolios. Therefore, ESG
shift portfolio weights from risky (low ESG) stocks to less factors significantly enhance the explanatory power of stand-
risky (high ESG) stocks. During expansions, however, we ard asset pricing models, indicating that a substantial part of
observe the opposite effect, when the SOC factor shows a the return variation can be traced to the ESG factors.
positive average return of 0.15 percent per month. For the This result is confirmed by a R2 decomposition using
CGV factor, average returns do not show large differences Shapley values (for example, Kruskal 1987).11 In unreported
between recessions and expansions. results, we decompose the R2 values of the decile portfolios
Overall, the results from Table 2 suggest that ESG risks and find that a substantial part of these can be allocated to
are associated with time-varying returns which, partly, relate the ESG factors, highlighting their relative importance. For
to market states. In line with our results, Dong et al. (2019) some of the decile portfolios, ESG factors show an even
report an outperformance of US equity funds overweighting higher relative importance than common factors such as
low ESG stocks compared to funds underweighting them by SMB, RMW or CMA.
about 2 percent annually (which reverses during the financial
crisis). Brammer et al. (2006) and Mollet and Ziegler (2014)
find the comparable results on the performance of portfolios Portfolio performance under consideration
formed using ESG ratings in Europe. of ESG

Explanatory power of ESG factors For investors, evidence for ESG-related stock return varia-
tion may raise the question of whether ESG risks are sys-
If there is ESG-related stock return variation, adding ESG fac- tematically related to portfolio performance. Therefore,
tors should improve the explanatory power of standard asset we investigate the relationship between ESG risks and
pricing models. We first sort all stocks into equally weighted risk–return profiles of equity portfolios.
decile portfolios based on their ESG exposures. The portfolios
are updated yearly and rebalanced on a monthly basis. We Performance of ESG quintile portfolios
then run time series regressions explaining the variation in
the portfolios’ monthly excess returns using the 6F-FFC and We compose equally weighted quintile portfolios based
the 9F-ESG model. We evaluate the added explanatory power on ESG ratings and ESG exposures and calculate risk and
of the ESG factors by examining the change in adj. R2 when return measures based on monthly returns (for example,
expanding the 6F-FFC to the 9F-ESG model.9 Auer and Schuhmacher 2016 and Galema et al. 2008). The
Table 3 shows coefficient estimates and changes in adj. portfolios are updated at the beginning of each year for our
R2.10 In Panel A, decile portfolios based on ENV exposures sample between 2003 and 2016.
show market beta coefficients about one. Adding ESG fac- Panel A of Table 4 presents mean excess returns of the
tors increases adj. R2 for all deciles on conventional sig- quintile portfolios. Concerning ENV ratings, we find that
nificance levels, indicating significant increases in explana- stocks in the top quintile (highest ratings) show an average
tory power. Large increases of more than 3 percent exhibit annual return of 8.77 percent which is 2.51 percent lower
the top decile (“green”, lowest ENV exposure of − 1.206) than the average return of the bottom quintile. However,
and the bottom decile (“brown”, highest ENV exposure of testing for equality, we do not find evidence for significant
1.232). Thus, both extreme deciles appear to be associated differences between the five mean returns. This is also true
with higher risks. Görgen et al. (2018) find the comparable for the mean returns of the quintile portfolios based on SOC
results for their carbon factor, which are somewhat more and CGV ratings. Compiling quintiles based on ESG expo-
pronounced in the highest carbon exposure decile. sures instead of ratings (lines four to six of Panel A) shows
The deciles based on SOC and CGV exposures in Pan- below-average returns for the majority of top and bottom
els B and C show the expected patterns. Similarly to the quintile portfolios across all three pillars. Again, we find
ENV exposures, we find that the ESG factors significantly that the mean returns of the quintile portfolios do not sig-
increase the adj. R2 for virtually all of the SOC and CGV nificantly differ from each other.12 Thus, there is no evidence

11
Shapely values account for the interplay between the individual
factors. For a deeper discussion of methods to decompose ­R2, see
9
We apply an F-test for nested models, as the 6F-FFC model is a Klein and Chow (2013).
subset of the 9F-ESG model. 12
Considering that quintile returns might be heteroskedastic or non-
10
Detailed results, including all coefficient estimates, are available normal, we also apply the Welch (1938) test and the nonparametric
upon request. test by Kruskal and Wallis (1952). The results remain robust.
Integrating sustainability risks in asset management: the role of ESG exposures and ESG ratings 59

Table 3  Explanatory power 9F-ESG model Explanatory power of ESG

of ESG factors based on ESG factors (%)
decile portfolios
𝛽 MKT 𝛽 ENV 𝛽 SOC 𝛽 CGV Adj. R2 Adj. R2 Δ adj. R2
6F-FFC 9F-ESG

Panel A: ENV deciles

Top (low exposure) 1.093*** − 1.206*** 0.481*** 0.078 86.03 89.06 3.02***
2 0.991*** − 0.711*** 0.421*** 0.045 92.74 94.24 1.50***
3 0.986*** − 0.221** 0.123 − 0.056 95.80 95.91 0.11*
4 0.913*** − 0.202*** 0.006 − 0.043 96.40 96.65 0.25***
5 0.993*** − 0.128** 0.193** 0.105 97.07 97.20 0.13**
6 0.932*** 0.182** 0.001 0.077 96.37 96.59 0.21***
7 1.001*** 0.125** − 0.010 0.043 97.15 97.20 0.14*
8 0.980*** 0.364*** − 0.122 0.109 96.32 96.90 0.58***
9 1.009*** 0.577*** − 0.348*** − 0.061 94.89 95.78 0.89***
Bottom (high exposure) 1.127*** 1.232*** − 0.759*** − 0.405*** 90.02 93.16 3.14***
Panel B: SOC deciles
Top (low exposure) 1.060*** 0.499*** − 1.106*** − 0.074 90.23 92.88 2.65***
2 1.004*** 0.262** − 0.557*** − 0.039 94.32 95.15 0.83***
3 0.962*** 0.039 − 0.294*** 0.057 96.43 96.80 0.37***
4 0.945*** 0.160** − 0.130* 0.171*** 96.80 96.97 0.17**
5 0.917*** 0.129* − 0.164** 0.079 96.69 96.77 0.08*
6 0.926*** − 0.050 0.031 − 0.089 97.15 97.14 − 0.01
7 0.958*** − 0.054 0.158** 0.144 96.70 96.85 0.15**
8 0.977*** − 0.198* 0.402*** − 0.091 95.60 96.12 0.51***
9 1.078*** − 0.347*** 0.649*** − 0.003 95.71 96.84 1.12***
Bottom (high exposure) 1.204*** − 0.382** 0.971*** − 0.226 88.22 89.89 1.67***
Panel C: CGV deciles
Top (low exposure) 1.064*** 0.030 0.149 − 1.539*** 83.28 87.78 4.51***
2 0.985*** 0.000 0.072 − 0.690*** 93.31 94.76 1.45***
3 0.948*** − 0.025 − 0.010 − 0.242** 95.69 95.86 0.17**
4 0.945*** 0.107 − 0.027 − 0.184** 96.20 96.30 0.10*
5 0.953*** 0.039 0.150** − 0.118 96.28 96.43 0.16**
6 0.962*** 0.061 − 0.036 0.164*** 97.04 97.11 0.07*
7 0.932*** 0.007 − 0.074 0.303*** 96.18 96.47 0.29***
8 1.032*** − 0.059 − 0.011 0.409*** 96.74 97.18 0.44***
9 1.056*** 0.045 − 0.196** 0.674*** 95.04 96.31 1.27***
Bottom (high exposure) 1.144*** − 0.106 − 0.075 1.072*** 91.06 93.10 2.04***

This table shows factor exposures and adj. R2 of ESG exposure decile portfolios. The results are calculated
based on monthly returns for the period from 2003 to 2016. Panels A, B and C report the results for the
ENV, SOC and CGV exposure deciles. Decile portfolios are updated at the beginning of each year accord-
ing to their current ESG ratings and ESG exposures. The last column investigates changes in adj. R2 when
the ESG factors are added to the 6F-FFC model (resulting in the 9F-ESG model). Significance levels are
calculated based on two-sided t-tests for exposures and one-sided F-tests for changes in adj. R2. Following
Newey and West (1987, 1994), we adjust standard errors for autocorrelation and heteroscedasticity. ***, **
and * indicate significance at the 1%, 5% and 10% level

for a persistent link between ESG risk exposures and average Investors usually optimize their risk–return profiles. As
stock returns.13 investors broadly consider ESG to improve risk manage-
ment, we explore whether ESG exposures are related to
13
portfolio risks. For each quintile portfolio, we calculate
Taking into account that returns on quintile portfolios could also
the volatilities of excess returns in Panel B. The rating
be driven by factor exposures, we calculate alphas from the 6F-FFC
model and again find no significant relationship between ESG expo- quintiles show very similar monthly return volatilities
sures and alphas. These results are available upon request.
60 B. Hübel, H. Scholz

Table 4  Risk and return measures of ESG quintile portfolios

Top 2 3 4 Bottom p value (%)
(H0: all quintiles equal)

Panel A: mean excess returns (%, p.a.)

Ratings
ENV 8.77 10.58 10.97 13.47 11.28 58.79
SOC 9.99 10.09 10.83 11.70 12.22 67.75
CGV 9.68 11.51 10.33 11.19 11.85 78.27
Exposures
ENV 9.69 11.62 10.93 12.15 10.71 86.27
SOC 10.54 11.69 11.67 11.47 9.76 90.42
CGV 12.19 10.86 11.41 11.22 9.34 71.93
Panel B: volatility of excess returns (%)
Ratings
ENV 5.14 4.92 4.82 5.05 4.99 93.69
SOC 4.78 5.00 4.93 5.21 5.02 86.31
CGV 4.86 4.99 5.03 4.90 5.11 96.87
Exposures
ENV 5.57 4.56 4.59 4.81 5.73 0.29
SOC 5.66 4.61 4.33 4.70 5.90 0.00
CGV 5.53 4.53 4.55 4.87 5.81 0.14
Panel C: 6F-FFC idiosyncratic volatility (%)
Ratings
ENV 1.70 1.72 2.18 2.59 1.76 6.69
SOC 2.14 2.07 2.77 2.47 2.03 32.21
CGV 2.07 2.59 1.67 2.29 2.06 37.66
Exposures
ENV 8.53 2.06 1.85 1.90 6.07 0.00
SOC 5.06 1.86 1.77 2.18 5.83 0.00
CGV 9.52 2.52 2.21 2.16 5.26 0.00
Panel D: 99% value-at-risk (%)
Ratings
ENV 14.62 15.34 14.31 16.30 16.27
SOC 14.34 14.90 14.39 17.05 15.55
CGV 15.60 14.84 15.14 14.80 15.37
Exposures
ENV 14.63 13.58 16.49 16.48 20.62
SOC 16.96 14.66 13.08 15.64 18.34
CGV 18.44 14.29 14.16 15.62 16.65

This table shows performance and risk measures of ESG quintile portfolios. Portfolios are updated and rebalanced at the beginning of every year
based on their current ESG rating and ESG exposures, respectively. All values are calculated based on monthly returns for the period from 2003
to 2016. The excess returns in Panel A are calculated as the difference between mean quintile portfolio return and the mean risk-free rate. The
volatilities in Panel B are calculated based on the monthly returns of each quintile portfolio. The idiosyncratic volatilities in Panel C are calcu-
lated using the 6F-FFC model. Panel D shows the historical monthly value-at-risk. The last column reports p values from significance tests on
equality of the quintiles. Mean excess returns in Panel A are tested based on a two-sided analysis of variance tests. Volatilities in Panels B and C
are evaluated using the test by Levene (1960)

of about 5 percent. Testing for equal variances based on the quintiles two, three and four. Testing for equality of
Levene (1960), we do not find any significant differences. all variances, we find p values below 1 percent indicat-
ESG ratings therefore seem to be unrelated to portfolio ing significant differences between the variances of the
risks. In contrast, the top and bottom ESG exposure port- quintile portfolio returns. However, this effect could be
folios are associated with higher volatilities compared to driven by differing factor exposures across the quintiles.
Integrating sustainability risks in asset management: the role of ESG exposures and ESG ratings 61

Fig. 2  Cumulative market- ENV rating quintile portfolios ENV exposures quintile portfolios
adjusted returns of ESG quintile
40% 40%
portfolios. This figure shows
the cumulative market-adjusted 30% 30%
monthly returns of the ESG 20% 20%
quintile portfolios based on
10% 10%
ESG ratings and ESG expo-
sures for the period from 2003 0% 0%
to 2016. The gray-shaded area -10% -10%
denotes the financial crisis from
May 2007 to February 2009 -20% -20%
-30% -30%
-40% -40%
2003 2005 2007 2009 2011 2013 2015 2003 2005 2007 2009 2011 2013 2015

ENV top ENV 2-4 ENV bottom ENV top ENV 2-4 ENV bottom

SOC rating quintile portfolios SOC exposures quintile portfolios

40% 40%
30% 30%
20% 20%
10% 10%
0% 0%
-10% -10%
-20% -20%
-30% -30%
-40% -40%
2003 2005 2007 2009 2011 2013 2015 2003 2005 2007 2009 2011 2013 2015

SOC top SOC 2-4 SOC bottom SOC top SOC 2-4 SOC bottom

CGV rating quintile portfolios CGV exposures quintile portfolios

40% 40%
30% 30%
20% 20%
10% 10%
0% 0%
-10% -10%
-20% -20%
-30% -30%
-40% -40%
2003 2005 2007 2009 2011 2013 2015 2003 2005 2007 2009 2011 2013 2015

CGV top CGV 2-4 CGV bottom CGV top CGV 2-4 CGV bottom

To control for common risk factors, we calculate the idi- risks (Panel D) as indicated by high historical value-at-risk
osyncratic volatility based on the 6F-FFC model (Panel values of up to 20.62 percent per month.
C). The results remain largely unchanged. Again, the top To sum up, we do not find evidence for a significant link
and bottom exposure quintiles show significantly higher between ESG risks and average stock returns. Still, it is
idiosyncratic risks than the remaining quintiles. These important to consider ESG in portfolio management from a
idiosyncratic risks can be largely explained by the three risk perspective. We find that very pronounced ESG expo-
ESG factors (see asset pricing tests in “Explanatory power sures are associated with high idiosyncratic risk. These risks
of ESG factors” section). In addition, the top and bottom can be explained by the ESG risk factors.
ESG exposure quintiles show relatively large downside
62 B. Hübel, H. Scholz

To illustrate the risks associated with very pronounced considerable differences in the country composition of the
ESG exposures, Fig. 2 shows the cumulative market-adjusted portfolios in Panel C.15 The rated portfolio is comprised,
returns of the ESG quintile portfolios. On the left-hand side, on average, of 30.99 percent of stocks invested from the
the market-adjusted returns of the top and bottom ESG rat- UK. This is 15.75 percentage points more than the unrated
ing quintiles (thick lines) show similar volatilities compared portfolio.
to the remaining quintiles (thin lines). In contrast, the top Besides differences due to average characteristics, dif-
and the bottom portfolios based on ESG exposures (right- ferences in the tails of the ESG exposure distributions of
hand side) show more volatile deviations from the market rated and unrated stocks might also affect portfolio composi-
return than the remaining quintiles. This effect is especially tion when including unrated stocks. To investigate this, we
pronounced for the ENV exposure portfolios during the estimate kernel densities based on the ESG exposures for
financial crisis (gray-shaded area). rated stocks and unrated stocks. We then test both densities
Quintile portfolios based on ESG ratings are per se indus- for equality using the nonparametric Kolmogorov–Smirnov
try-adjusted due to the industry adjustment in the ESG rat- test (Frank and Massey 1951). In the unreported results,
ing methodology. ESG exposure quintile portfolios, how- we find that the ESG exposures of rated and unrated stocks
ever, may involve deviations from market industry weights differ significantly from each other.16 Although economi-
that affect risk and returns of the portfolios. Table 5 shows cally small, this difference could affect risk and returns of
the average industry weights of the ESG exposure quintile ESG portfolios. We therefore split each quintile portfolio
portfolios. (see Table 4) into two sub-portfolios containing rated and
Indeed, the industry weights differ substantially across unrated stocks, respectively.
the quintiles. For instance, the bottom ENV exposure quin- Risk and return measures for the sub-portfolios are shown
tile (high ENV risks) shows significantly lower weights on in Table 7. As shown in Panel A, there seem to be differ-
Financials and higher weights on Oil and Gas compared to ences between the mean excess returns within the top and
the top ENV exposure quintile. This appears to be reason- bottom quintiles based on ENV and SOC exposures, but not
able, as operating in the Oil and Gas industry is usually based on CGV exposures. Regarding volatilities in Panel B,
associated with extensive resource use and greenhouse gas there is no evidence for any systematic difference between
emissions.14 rated and unrated sub-portfolios. Panel C shows that the cor-
relations between the returns of the rated and unrated sub-
Rated versus unrated stocks portfolios range between 0.89 and 0.95. All correlations are
significantly positive on the 1% level. This suggests that the
As stock returns are broadly available, portfolios based on returns do not only share similar volatilities, but also highly
ESG exposures include unrated stocks which cannot be con- positive correlations.
sidered for portfolios based on ESG ratings. Therefore, we To sum up, we find differences in characteristics and risk
investigate whether the inclusion of unrated stocks affects exposures between rated and unrated stocks. These result in
our results. First, we compare stock characteristics as well as slight differences in average returns but not in volatilities. As
industry and country exposures of rated and unrated stocks. our study focuses on portfolio risks, we conclude that there
To do so, we divide the sample into two portfolios contain- is no clear evidence for unrated stocks systematically affect-
ing the rated stocks and the unrated stocks and update the ing our previous results. However, a deeper discussion of the
equal-weighted portfolios on a yearly basis. differences between the mean returns of rated and unrated
Panel A of Table 6 compares stock fundamentals of the stocks could be considered for future research.
rated and the unrated portfolio. Unrated stocks are on aver-
age more than four times smaller than rated stocks. Also, ESG screenings from an investor’s perspective
unrated stocks have lower book-to-market ratios. Our results
are in line with the findings of Galema et al. (2008), for the From an investor’s point of view, avoiding stocks with pro-
US market. Therefore, disregarding unrated stocks could nounced ESG risks reduces portfolio risks. However, avoid-
lead to a systematic disregard of small growth stocks. ing stocks with high ESG risks also means a smaller invest-
In Panel B, we compare the industry compositions of ment universe. To explore whether reducing the investment
both portfolios. Although we find some statistically signifi- universe has an impact on performance, we compare the
cant differences, the overall industry composition appears performance of widely used ESG screenings with a passive
similar from an economic standpoint. However, we find

15
For brevity, we report significant results only. Detailed results are
14
Compare the discussion of the differences between ESG ratings available upon request.
16
and ESG exposures in “ESG exposures versus ESG ratings” section. Results are available upon request.
Integrating sustainability risks in asset management: the role of ESG exposures and ESG ratings 63

Table 5  Industry weights of ESG exposure quintile portfolios (%)

Top 2 3 4 Bottom p value (%)
(H0: all quintiles equal)

ENV exposure
Basic materials 5.2 4.8 6.3 7.2 10.0 0.00
Consumer goods 9.2 10.6 11.2 11.1 12.2 0.00
Consumer services 16.3 14.9 14.3 14.0 11.8 0.00
Financials 29.3 23.7 22.6 22.7 15.3 0.00
Health care 5.1 6.7 6.6 6.4 6.0 0.31
Industrials 21.2 22.9 22.9 23.3 23.3 0.01
Oil and gas 3.5 3.3 3.7 3.9 8.7 0.00
Technology 5.1 5.1 5.3 5.1 5.8 0.29
Telecommunications 2.9 3.2 2.6 2.2 2.5 0.00
Utilities 2.9 4.8 4.4 4.4 4.3 0.02
SOC exposure
Basic materials 10.3 6.4 6.1 5.5 5.2 0.00
Consumer goods 11.4 12.6 11.7 10.8 7.9 0.00
Consumer services 14.8 15.2 13.8 12.8 14.6 0.37
Financials 14.1 18.7 22.7 27.5 30.6 0.00
Health care 5.5 6.2 6.5 6.2 6.3 0.25
Industrials 22.3 23.9 23.8 22.5 21.1 0.03
Oil and gas 7.1 4.6 3.6 3.6 3.9 0.00
Technology 7.0 5.2 4.3 4.7 5.2 0.00
Telecommunications 2.1 1.9 3.1 3.1 3.3 0.00
Utilities 5.4 5.2 4.3 3.4 2.0 0.00
CGV exposure
Basic materials 8.1 6.8 6.6 6.1 6.1 0.00
Consumer goods 7.0 8.7 10.2 12.1 16.4 0.00
Consumer services 14.1 15.7 14.5 13.5 13.5 0.00
Financials 22.7 25.4 24.0 22.4 19.0 0.00
Health care 5.6 6.8 7.0 6.3 5.2 0.38
Industrials 20.0 21.0 21.6 24.9 26.1 0.00
Oil and gas 10.2 4.4 3.7 2.7 3.0 0.00
Technology 6.9 4.7 4.7 4.8 5.3 0.00
Telecommunications 2.5 2.7 3.2 2.9 2.5 56.91
Utilities 3.2 3.8 4.7 4.9 3.8 0.00

This table shows average industry weights of quintile portfolios based on ESG exposures. The time series averages of the mean industry weights
are calculated using the Industry Classification Benchmark codes from EIKON. Portfolios are updated and rebalanced at the beginning of every
year based on their current ESG exposures. The last column reports p values from significance tests on equality of the quintiles based on a two-
sided analysis of variance tests

investment in the market (Renneboog et al. 2008). To do from the stock universe. We apply cutoff rates between 10
so, we look at the performance of several equity portfolios percent and 90 percent. To assess risk-adjusted performance,
screened by ESG ratings and exposures, applying common we calculate mean excess returns and alphas based on the
positive and negative screening procedures with different 6F-FFC for each of the nine portfolios per ESG pillar.
cutoffs. More precisely, we rank all stocks based on their The first column of Table 8 shows that positive screen-
ESG ratings or ESG exposures at the end of every year end. ings based on ENV ratings are associated with lower mean
Then, we form equal-weighted portfolios excluding the excess returns than the equally weighted benchmark port-
stocks with the lowest ranks (negative screening) or only folio consisting of all stocks (100 percent cutoff). The
including stocks with the highest ranks (positive screening). underperformance is highly significant and ranges between
The cutoff rate specifies the proportion of stocks selected 2.98 percent p.a. (10 percent cutoff) and 1.11 percent p.a.
64 B. Hübel, H. Scholz

Table 6  Characteristics of the All stocks Rated portfolio Unrated portfolio Rated–unrated
rated portfolio and unrated
portfolio Panel A: stock fundamentals
# of stocks 1024 756 269 487***
Market capitalization (m€) 7725 9507 2162 7344***
Dividend yield (%) 0.69 0.66 0.81 0.21
Book-to-market 3.39 3.44 3.23 − 0.15***
Mean excess return (%) 0.91 0.91 0.91 0.00
SD of excess returns (%) 4.92 4.94 5.00 0.06
Panel B: industry weights (%)
Basic materials 6.55 7.18 4.52 2.67***
Consumer goods 10.84 11.06 10.64 0.42
Consumer services 14.48 14.95 12.87 2.08***
Financials 22.63 21.69 25.63 − 3.94***
Health care 6.04 6.09 6.05 0.03
Industrials 22.53 22.03 24.56 − 2.53
Oil and gas 4.55 4.84 3.7 1.14*
Technology 5.52 4.79 6.85 − 2.06
Telecommunications 2.87 3.09 2.1 0.99
Utilities 3.98 4.28 3.23 1.05
Panel C: country weights (%)
Denmark 2.70 3.06 1.69 1.37*
Italy 7.88 6.36 12.64 − 6.28***
Luxembourg 0.50 0.38 1.08 − 0.71***
United Kingdom 27.67 30.99 15.24 15.75***

This table reports characteristics of the portfolios containing rated and unrated stocks. Panel A shows time
series averages of various stock fundamentals. To calculate excess returns, we use the 1-month EURIBOR.
Panel B reports the time series averages of the mean industry weights using the Industry Classification
Benchmark codes from EIKON. Panel C shows the average country weights using the geographical clas-
sification codes from EIKON. For brevity in Panel C, we report the significant results only. The detailed
results are available upon request. ***, ** and * indicate significance at the 1%, 5% and 10% level

(50 percent cutoff). This finding is in line with Trinks and The results for the CGV pillar show that negative and
Scholtens (2017) who investigate negative ESG screenings positive screenings based on CGV ratings slightly decrease
based on a global sample. The underperformance weakens excess returns, while for the 6F-FFC alpha, there seems to be
but remains significant when we look at the alphas in Panel no link between CGV ratings and risk-adjusted performance.
B. Ranking stocks based on ESG exposures, we find a simi- As to positive screenings based on CGV exposures, we find
larly deteriorating performance when positive screenings are that returns mostly increase. However, neither returns nor
applied. Notably, none of the t-values indicates a significant alphas based on the 6F-FFC model show significant results.
difference between benchmark return and ENV-exposure- We therefore find that CGV screenings do not have an
screened portfolio returns. In line with our previous results, impact on performance.
this points to an insignificant relationship between ENV In summary, we derive three main results regarding
exposures and financial performance. We thus do not find the performance impact of ESG screenings. First, positive
any significant performance impact due to screenings based ENV screenings are associated with a systematic sacrifice
on ENV exposures. of financial performance when applied on ENV ratings, but
The results for screenings based on SOC ratings and do not show an impact on performance when applied based
exposures show analogous results. Similar to ENV ratings, on ENV exposures. Second, portfolio returns decrease when
we find significantly lower mean excess returns than the mar- investors select firms with the highest SOC ratings. SOC
ket (approx. 1 percent p.a.) for positive screenings based on exposures, however, are virtually unrelated to financial per-
SOC ratings. The underperformance is, however, not robust formance. Third, we do not find any significant relationship
to controlling for common risk factors, as 6F-FFC alphas are between CGV ratings or exposures and financial perfor-
insignificant. Applying screenings based on SOC exposures mance. Our results thus indicate that investors can reduce
does not indicate a significant impact on performance.
Integrating sustainability risks in asset management: the role of ESG exposures and ESG ratings 65

Table 7  Risk and return of rated and unrated stocks within ESG exposure quintile portfolios
Top 2 3 4 Bottom

Panel A: mean annualized excess returns of quintile portfolios (%)

ENV
All stocks 9.69 11.62 10.93 12.15 10.71
Rated 8.49 11.62 11.18 11.74 11.52
Unrated 12.52 12.56 12.44 14.00 10.03
Rated–unrated − 4.02 − 0.94 − 1.26 − 2.26 1.49
t-stat (− 1.685) (0.413) (0.105) (− 0.678) (1.666)
SOC
All stocks 10.54 11.69 11.67 11.47 9.76
Rated 11.03 11.45 11.74 11.40 9.31
Unrated 9.22 12.11 11.84 13.46 11.59
Rated–unrated 1.81 − 0.66 − 0.10 − 2.07 − 2.69
t-stat (1.808) (0.441) (0.889) (−0.570) (− 1.939)
CGV
All stocks 12.19 10.86 11.41 11.22 9.34
Rated 11.58 10.83 11.24 10.96 9.86
Unrated 12.76 11.82 12.67 12.59 10.94
Rated–unrated − 1.17 − 0.99 − 1.43 − 1.62 − 1.08
t-stat (0.040) (0.191) (−0.153) (−0.292) (0.124)
Panel B: volatilities of excess returns of quintile portfolios (%)
ENV
All stocks 5.57 4.56 4.59 4.81 5.73
Rated 5.56 4.58 4.56 4.93 5.74
Unrated 5.69 4.79 4.77 4.62 6.17
Rated–unrated − 0.13 − 0.21 − 0.20 0.31 − 0.43
p value (%) 76.43 57.03 57.11 40.65 35.37
SOC
All stocks 5.66 4.61 4.33 4.70 5.90
Rated 5.73 4.58 4.35 4.73 5.99
Unrated 6.12 5.22 4.56 4.86 5.96
Rated–unrated − 0.39 − 0.64 − 0.21 − 0.12 0.02
p value (%) 39.59 9.44 53.55 74.32 95.85
CGV
All stocks 5.53 4.53 4.55 4.87 5.81
Rated 5.57 4.58 4.51 4.88 5.88
Unrated 6.10 4.95 4.65 5.09 5.91
Rated–unrated − 0.53 − 0.37 − 0.14 − 0.22 − 0.04
p value (%) 23.94 31.33 70.07 57.30 93.70
Panel C: correlations of excess returns of rated and unrated quintile portfolios (%)
ENV
Correlation 0.92 0.92 0.95 0.94 0.91
p value (%) 0.00 0.00 0.00 0.00 0.00
SOC
Correlation 0.91 0.91 0.92 0.94 0.93
p value (%) 0.00 0.00 0.00 0.00 0.00
CGV
Correlation 0.89 0.90 0.94 0.94 0.93
p value (%) 0.00 0.00 0.00 0.00 0.00

This table shows risk and return measures for ESG exposure quintile portfolios. Each portfolio is decomposed in sub-portfolios based on rated and unrated
stocks. All values are calculated based on monthly returns of quintile portfolios for the period from 2003 to 2016. Panel A shows mean annualized excess
returns, Panel B volatilities of excess returns and Panel C correlations of excess returns of rated and unrated sub-portfolios based on ESG exposures. t-sta-
tistics are reported in parentheses. Bold values indicate t-values significant at the 10% level. Following Newey and West (1987, 1994), we adjust standard
errors for autocorrelation and heteroskedasticity. p values in Panel B are calculated from significance tests on equality of the volatilities of the rated and
unrated sub-portfolio using the test by Levene (1960). p values in Panel C stem from testing correlations being significantly different from zero
66 B. Hübel, H. Scholz

Table 8  Performance of ESG-screened portfolios

This table reports performance measures for ESG-screened portfolios. We sort stocks at the beginning of each year based on their current ESG
rating and ESG exposures, respectively. Then, different cutoff rates are applied to compose portfolios. The cutoff rate specifies the proportion
of stocks selected for each portfolio. We update and rebalance the equally weighted portfolios at the beginning of every year. All values are
calculated based on monthly returns for the period from 2003 to 2016. The excess returns in Panel A are calculated in excess of the 1-month
EURIBOR. The alphas in Panel B are calculated based on the 6F-FFC model applying the equally weighted benchmark portfolio consisting of
all stocks (100 percent cutoff) as market factor. Data bars indicate the deviation from the equally weighted benchmark portfolio consisting of all
stocks (100 percent cutoff). All portfolio measures are annualized. t-statistics are reported in parentheses. Bold t-values indicate significance at
the 10% level. Following Newey and West (1987, 1994), we adjust standard errors for autocorrelation and heteroscedasticity

ESG risk exposures without sacrificing financial perfor- performance of stocks without ESG ratings has an impact
mance, regardless of the ESG pillar. on these conclusions.
As the rising sustainability awareness affects firm values
(due to changes in stakeholder behavior, ESG regulation,
Conclusion etc.), market pricing of stocks can be expected to reflect
firms’ ESG risks. Against this background, we develop a
The increasing sustainability awareness of governments, multifactor model to measure ESG exposures of firms or
regulatory authorities and customers encourages firms to portfolios. In particular, we introduce three environmental,
adopt sustainable business practices. This rise in awareness social and governance factors with which to measure the
regarding sustainability is changing the economic environ- stock market’s time-varying perception and valuation of
ment and involves profound changes on capital markets. ESG risks.
Because ESG risks are hardly diversifiable, it has become Based on a comprehensive sample of European stocks
critically important for investors to understand the ESG pro- between 2003 and 2016, we derive the four main empirical
file of the firms and portfolios they are investing in. Thus, results. First, comparing ESG ratings and ESG exposures
comparable and reliable ESG data are of crucial importance of firms shows a positive but not perfect relationship. These
in investment decision-making and sophisticated risk man- differences between ratings and exposures may be due to
agement. In this paper, we propose applying return-based the industry benchmarking of ESG ratings and to ESG risks
ESG exposures to measure and integrate ESG risks into which are not assigned to firms by ESG rating agencies.
asset management. Second, firms with low environmental ratings outperformed
We contribute to the existing ESG literature in three firms with high environmental ratings, as indicated by the
ways. First, we explore three ESG risk factors in order to ENV factor showing a significant and positive mean annual
capture ESG risk exposures of firms. Second, we draw con- return of about 3 percent. Third, in line with a “flight to qual-
clusions regarding the relationship between ESG factors and ity” effect during crisis periods, highly social firms outper-
performance by composing portfolios based on ESG rat- form less social firms during crises, as indicated by signifi-
ings and ESG exposures. Third, we investigate whether the cant negative returns of the SOC factor during recessions.
Integrating sustainability risks in asset management: the role of ESG exposures and ESG ratings 67

Fourth, adding ESG factors significantly enhances the Martin Rohleder, Stephen Satchell, Ingrid Tierens, Sebastian Utz,
explanatory power of standard asset pricing models with Abhishek Varma, Krishna Vyas, Martin Wallmeier, Marco Wilkens,
Maximilian Wimmer and the participants of the SWFA Conference
regard to ESG decile portfolios. However, there is no evi- 2018 in Albuquerque, the Green Summit 2018 in Vaduz, the Summer
dence for a systematic ESG-related return premium or dis- Conference on Financial Implications of Sustainability and Corporate
count. Whereas we find that very pronounced ESG expo- Social Responsibility 2018 in Nice, the International Ph.D. Seminar
sures are related to higher portfolio risks, investors can still 2018 in Hagen and the Inquire Europe 2018 autumn seminar in Buda-
pest. The paper received the Best Doctoral Student Paper in Invest-
compose portfolios with lower ESG risk exposures while ments Award at the 2018 SWFA Annual Meeting and the Inquire Prize
keeping risk-adjusted performance virtually unchanged. for the best paper presentation on „ESG and portfolio construction“
Investigating risk and return properties of portfolios based at the 2018 Inquire Europe autumn seminar. A former version of this
on ESG ratings and ESG exposures enables us to derive manuscript was entitled “Performance of socially responsible port-
folios: The role of CSR exposures and CSR ratings”. All remaining
implications from an investor’s perspective. Whereas stocks errors are our own.
with the best ENV ratings underperform the market return
by about 2 percent annually, SOC and CGV ratings appear
to be unrelated to portfolio returns. This could be attributed
to an overvaluation of stocks with high ENV ratings due References
to a rising awareness of environmental risks among inves-
tors. Composing portfolios based on ESG exposures, we do Ameer, R. 2016. Usefulness of ASSET4 ESG for Socially Respon-
not find that ESG exposures have a substantial impact on sible Investing in a Small Economy. Journal of Investing 25(2):
portfolio returns. However, portfolios with pronounced ESG 125–135.
Amiraslani, H., K.V. Lins, H. Servaes, and A.M. Tamayo. 2019. The
exposures show higher risks as the remaining stocks in the Bond Market Benefits of Corporate Social Capital. Brussels:
market. A large proportion of these risks can be explained European Corporate Governance Institute.
by the ESG factors. Also, we find that investors can compose Aon Benfield, 2017. Weather, Climate & Catastrophe Insight. http://
portfolios with lower ESG risk exposures while keeping thoug​htlea​dersh​ip.aonbe​nfiel​d.com/Docum​ents/20180​124-ab-
if-annua​l -repor ​t -weath​e r-clima​t e-2017.pdf. Accessed 1 Mar
risk-adjusted performance virtually unchanged. Calculating 2019.
ESG exposures thus enables investors to avoid ESG risks Auer, B.R. 2016. Do Socially Responsible Investment Policies Add
without sacrificing financial performance. or Destroy European Stock Portfolio Value? Journal of Business
Our study relies on ESG ratings provided by Thomson Ethics 135(2): 381–397.
Auer, B.R., and F. Schuhmacher. 2016. Do Socially (Ir)responsible
Reuters EIKON. While Thomson Reuters is a major data Investments Pay? New Evidence from International ESG Data.
provider widely used by investors and academics, future Quarterly Review of Economics and Finance 59: 51–62.
research would likely benefit from calculating ESG fac- Barth, F., C. Eckert, N. Gatzert, and H. Scholz. 2019a. Spillover
tors based on data by other ESG rating agencies to evaluate Effects from the Volkswagen Emissions Scandal: An Analysis
of Stock, Corporate Bond, and Credit Default Swap Markets.
the impact of different methodologies across ESG rating Nürnberg: Friedrich-Alexander-Universität Erlangen-Nürnberg.
agencies. Since we construct the three ESG factors as zero- Barth, F., B. Hübel, and H. Scholz. 2019b. ESG and Corporate
investment portfolios based on diversified top and bottom Credit Spreads. Nürnberg: Friedrich-Alexander-Universität
tercile portfolios, we expect that manageable differences in Erlangen-Nürnberg.
Borgers, A., J. Derwall, K. Koedijk, and J. ter Horst. 2015. Do Social
ESG ratings should not result in substantial differences in Factors Influence Investment Behavior and Performance? Evi-
the composition of the tercile portfolios. In this context, we dence from Mutual Fund Holdings. Journal of Banking &
note that the correlations between the ESG firm ratings of Finance 60: 112–126.
different rating agencies are (not perfectly) but usually sig- Brammer, S.J., C. Brooks, and S. Pavelin. 2006. Corporate Social
Performance and Stock Return UK Evidence form Disaggregate
nificantly positive (see, for example, Chatterji et al. 2015; Measures. Financial Management 35(3): 97–116.
Dorfleitner et al. 2015). Carhart, M. 1997. On Persistence in Mutual Fund Performance.
In summary, taking ESG risk into account when man- Journal of Finance 52(1): 57–82.
aging equity portfolios enables investors to better assess CFA Institute. 2017. Environmental, Social and Governance Sur-
vey. https​: //www.cfain​stitu​t e.org/en/resea​rch/surve​y-repor ​t s/
the ESG risk exposures of their portfolios solely based on esg-surve​y-2017. Accessed 1 Mar 2019.
the high informational content of stock returns. In addi- Chatterji, A., R. Durand, D. Levine, and S. Touboul. 2015. Do Rat-
tion, investors can easily add ESG exposures into their risk ings of Firms Converge? Implications for Managers, Investors
management systems. Strategically managing these ESG and Strategy Researchers. Strategic Management Journal 37(8):
1597–1614.
risks may result in potential benefits for investors. Chava, S. 2014. Environmental Externalities and Cost of Capital.
Management Science 60(9): 2223–2247.
Acknowledgements We are grateful for helpful comments and sugges- Derwall, J., N. Guenster, R. Bauer, and K. Koedijk. 2005. The Eco-
tions from Matthias Bank, Rainer Baule, David Buckle, Ulf Erlandsson, Efficiency Premium Puzzle. Financial Analysts Journal 61(2):
Xavier Gerard, Binam Ghimre, Bart van der Grient, Dennis Huber, 51–63.
Marielle de Jong, Antoine Mandel, Jason MacQueen, Sarah McCarthy,
Zoltán Nagy, Martin Nerlinger, Satya Pradhuman, Cyrus Ramezani,
68 B. Hübel, H. Scholz

Dong, X., S. Feng, S. Parida, and Z. Wang. 2019. Corporate Social Kruskal, W. 1987. Relative Importance by Averaging Over Orderings.
Responsibility Exposure and Performance of Mutual Funds. American Statistician 41(1): 6–10.
Journal of Investing 28(2): 53–65. Kruskal, W.H., and W.A. Wallis. 1952. Use of Ranks in One-Criterion
Dorfleitner, G., G. Halbritter, and M. Nguyen. 2015. Measuring the Variance Analysis. Journal of the American statistical Association
Level and Risk of Corporate Responsibility—An Empirical 47(260): 583–621.
Comparison of Different ESG Rating Approaches. Journal of Levene, H. 1960. Robust testes for equality of variances. In Contribu-
Asset Management 16(7): 450–466. tions to Probability and Statistics, edited by I. Olkin, 278–292.
Dorfleitner, G., S. Utz, and M. Wimmer. 2018. Patience Pays Off— Stanford Univ. Press, Palo Alto, CA.
Corporate Social Responsibility and Long-Term Stock Returns. Lins, K.V., H. Servaes, and A. Tamayo. 2017. Social Capital, Trust,
Journal of Sustainable Finance & Investment 8(2): 132–157. and Firm Performance: The Value of Corporate Social Respon-
Eccles, R.G., I. Ioannou, and G. Serafeim. 2014. The Impact of Cor- sibility During the Financial Crisis. Journal of Finance 72(4):
porate Sustainability on Organizational Processes and Perfor- 1785–1824.
mance. Management Science 60(11): 2835–2857. Mollet, J.C., and A. Ziegler. 2014. Socially Responsible Investing
Edmans, A. 2011. Does the Stock Market Fully Value Intangibles? and Stock Performance: New Empirical Evidence for the US and
Employee Satisfaction and Equity Prices. Journal of Financial European Stock Markets. Review of Financial Economics 23(4):
Economics 101(3): 621–640. 208–216.
El Ghoul, S., O. Guedhami, C.C.Y. Kwok, and D.R. Mishra. 2011. Newey, W.K., and K.D. West. 1987. A Simple, Positive Semi-Definite,
Does Corporate Social Responsibility Affect the Cost of Capi- Heteroskedasticity and Autocorrelation Consistent Covariance
tal? Journal of Banking & Finance 35(9): 2388–2406. Matrix. Econometrica 55(3): 703–708.
El Ghoul, S., and A. Karoui. 2017. Does Corporate Social Respon- Newey, W.K., and K.D. West. 1994. Automatic Lag Selection in
sibility Affect Mutual Fund Performance and Flows? Journal Covariance Matrix Estimation. Review of Economic Studies 61(4):
of Banking & Finance 77: 53–63. 631–653.
European Commission. 2018. Sustainable Finance: Making the Nofsinger, J., and A. Varma. 2014. Socially Responsible Funds and
Financial Sector a Powerful Actor in Fighting Climate Change, Market Crises. Journal of Banking & Finance 48: 180–193.
24 May, http://europ​a .eu/rapid​/ press​- relea​s e_IP-18-3729_ Parguel, B., F. Benoît-Moreau, and F. Larceneux. 2011. How Sustain-
en.htm. Accessed 1 Mar 2019. ability Ratings Might Deter ‘Greenwashing’: A Closer Look at
Fama, E.F., and K.R. French. 1993. Common Risk Factors in the Ethical Corporate Communication. Journal of Business Ethics
Returns on Stocks and Bonds. Journal of Financial Economics 102(1): 15.
33(1): 3–56. Renneboog, L., J. Ter Horst, and C. Zhang. 2008. Socially Responsi-
Fama, E.F., and K.R. French. 2015. A Five-Factor Asset Pricing ble Investments: Institutional Aspects, Performance, and Investor
Model. Journal of Financial Economics 116(1): 1–22. Behavior. Journal of Banking & Finance 32(9): 1723–1742.
Frank, J., and J. Massey. 1951. The Kolmogorov–Smirnov Test for Scholtens, B. 2017. Why Finance Should Care About Ecology. Trends
Goodness of Fit. Journal of the American Statistical Association in Ecology & Evolution 32(7): 500–505.
46(253): 68–78. Statman, M. 2016. Classifying and Measuring the Performance of
Galema, R., A. Plantinga, and B. Scholtens. 2008. The Stocks at Stake: Socially Responsible Mutual Funds. Practical Applications 4(3):
Return and Risk in Socially Responsible Investment. Journal of 43–46.
Banking & Finance 32(12): 2646–2654. Statman, M., and D. Glushkov. 2016. Classifying and Measuring the
Gloßner, S. 2018. The Price of Ignoring ESG Risks. Eichstaett-Ingol- Performance of Socially Responsible Mutual Funds. Journal of
stadt: Catholic University of Eichstaett-Ingolstadt. Portfolio Management 42(2): 140–151.
Glück, M., B. Hübel, and H. Scholz. 2019. Currency Conversion of Stellner, C., C. Klein, and B. Zwergel. 2015. Corporate Social Respon-
Fama/French Factors: How and Why?. Nürnberg: Friedrich- sibility and Eurozone Corporate Bonds: The Moderating Role
Alexander-Universität Erlangen-Nürnberg. of Country Sustainability. Journal of Banking & Finance 59:
Görgen, M., A. Jacob, M. Nerlinger, R. Riordan, M. Rohleder, and M. 538–549.
Wilkens. 2018. Carbon Risk. Augsburg: University of Augsburg. Trinks, P.J., and B. Scholtens. 2017. The Opportunity Cost of Negative
Global Sustainable Investment Alliance GSIA. 2017. Global Sustaina- Screening in Socially Responsible Investing. Journal of Business
ble Investment Review 2016. 28 March 2017, http://www.gsi-allia​ Ethics 140(2): 193–208.
nce.org/wp-conte​nt/uploa​ds/2017/03/GSIR_Revie​w2016​.F.pdf. United Nations. 2015. Transforming our World: The 2030 Agenda for
Accessed 1 Mar 2019. Sustainable Development. 21 October, Resolution A/RES/70/1.
Henke, H.M. 2016. The Effect of Social Screening on Bond Mutual Utz, S., and M. Wimmer. 2014. Are They Any Good at All? A Finan-
Fund Performance. Journal of Banking & Finance 67: 69–84. cial and Ethical Analysis of Socially Responsible Mutual Funds.
Henriksson, R., J. Livnat, P. Pfeifer, and M. Stumpp. 2019. Integrating Journal of Asset Management 15(1): 72–82.
ESG in Portfolio Construction. Journal of Portfolio Management Welch, B.L. 1938. The Significance of the Difference Between Two
45(4): 67–81. Means When the Population Variances are Unequal. Biometrika
Hong, H., and M. Kacperczyk. 2009. The Price of Sin: The Effects 29: 350–362.
of Social Norms on Markets. Journal of Financial Economics Wimmer, M. 2013. ESG-Persistence in Socially Responsible Mutual
93(1): 15–36. Funds. Journal of Management and Sustainability 3(1): 9–15.
Ince, O.S., and R.B. Porter. 2006. Individual Equity Return Data from
Thomson Datastream: Handle With Care! Journal of Financial Publisher’s Note Springer Nature remains neutral with regard to
Research 29(4): 463–479. jurisdictional claims in published maps and institutional affiliations.
Jin, I. 2018. Is ESG a Systematic Risk Factor for US Equity Mutual
Funds? Journal of Sustainable Finance & Investment 8(1): 72–93.
Kempf, A., and P. Osthoff. 2008. SRI Funds: Nomen est Omen. Journal Benjamin Hübel is a Ph.D. candidate and lecturer at the Chair of
of Business Finance and Accounting 35(9–10): 1276–1294. Finance and Banking, Friedrich-Alexander-Universität Erlangen-Nürn-
Klein, R.F., and V.K. Chow. 2013. Orthogonalized Factors and Sys- berg, Germany, where he also graduated in business administration. His
tematic Risk Decomposition. Quarterly Review of Economics and research focuses on empirical asset pricing, mutual fund performance
Finance 53(2): 175–187. and sustainable finance.
Integrating sustainability risks in asset management: the role of ESG exposures and ESG ratings 69

Hendrik Scholz is the Chair of Finance and Banking at Friedrich-Alex- analysis, asset management, sustainable finance, structured financial
ander-Universität Erlangen-Nürnberg, Germany. He graduated from products and bank management. He has authored numerous articles in
University of Göttingen, Germany, where he also finished his Ph.D. finance journals and is a member of the Southern Finance Association,
Afterward, he finished his habilitation at the Catholic University of the German Finance Association and the German Academic Associa-
Eichstätt-Ingolstadt, Germany. His research focuses on performance tion for Business Research.
Reproduced with permission of copyright owner. Further reproduction
prohibited without permission.