15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2175 by University Of Embu, Wiley Online Library on [30/01/2023].

See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Received: 27 January 2020 Revised: 20 November 2020 Accepted: 30 November 2020
DOI: 10.1002/csr.2175

RESEARCH ARTICLE

The level of sustainability and mutual fund performance

in Europe: An empirical analysis using ESG ratings

Guido Abate | Ignazio Basile | Pierpaolo Ferrari

Department of Economics and Management,

University of Brescia, Brescia, Italy Abstract
Over recent years, investors' attention on the environment, social responsibility, and
Correspondence
Guido Abate, Department of Economics and governance (ESG) has been growing. At the same time, managers, investors, and
Management, University of Brescia, Brescia, regulators are interested in ascertaining whether mutual funds that invest in ESG-
Italy.
Email: [email protected] compliant assets perform better than those with a low ESG commitment. The
sustainability of funds' portfolios can be measured by ESG ratings, a measure of the
financially material ESG factors of the securities held by a fund. Our study therefore
aims to verify whether funds with high ESG ratings outperform funds with low ESG
ratings, considering the risks taken, including higher moments, and costs borne by
investors. Our analysis is carried out on a sample of 634 European mutual funds.
By using data envelopment analysis, it provides evidence of the superior efficiency of
funds investing in high ESG-rated securities.

KEYWORDS
data envelopment analysis, ESG ratings, mutual funds, risk-adjusted performance, sustainable
investments

JEL CLASSIFICATION
G10; G11; G15

1 | I N T RO DU CT I O N third section describes the dataset and methodology of the empirical

study, which employs data envelopment analysis (DEA), a non-
Over recent years, environmental, social, and governance (ESG) issues parametric method derived from operational research. The fourth
have received increasing attention worldwide (Amel-Zadeh & section details the empirical analysis, aimed at identifying the pres-
Serafeim, 2018; Engle III et al., 2020). In this context, asset managers ence of significant differences between the performance of funds
integrate these concerns into their investment policies (van Duuren with portfolios characterized by high ESG ratings and that of
et al., 2016), but the degree of sustainability of mutual funds varies. funds that make investments in assets with low ESG ratings. The fifth
This study comparatively evaluates the efficiency of mutual funds section concludes and provides an interpretation of the results.
characterized by different levels of sustainability, as measured by the
ESG rating of their portfolios, to investigate the potential presence of
superior risk-adjusted performances by sustainable funds. The aim 2 | LI T E RA T U R E RE V I E W
of this analysis, therefore, is to provide a rational framework for the
stakeholders involved in sustainable finance interested in funds' per- This section analyzes the fundamentals underlying the performance of
formance evaluation and information disclosure on asset allocation. sustainable mutual funds. We review the two main branches of the
The remainder of this paper is organized as follows. The second available scientific literature. The first focuses on the relationship
section reviews the scientific literature focusing on the performance between sustainable corporate policies and the economic returns
of sustainable investments and the factors to which it is subject. The achieved by the firms that adopt them; the second compares the

1446 © 2021 ERP Environment and John Wiley & Sons Ltd. wileyonlinelibrary.com/journal/csr Corp Soc Responsib Environ Manag. 2021;28:1446–1455.
 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2175 by University Of Embu, Wiley Online Library on [30/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABATE ET AL. 1447

performance of sustainable investment strategies with that of tradi- most relevant literature (Clark et al., 2015; United Nations Environ-
tional strategies to find any deviations between the two, defined as ment Program Finance Initiative & Mercer, 2007). Empirical analyses
ethical sacrifice (Becchetti et al., 2015). conducted at both the academic and the operational levels show an
From a theoretical point of view, the first branch of research is evolution over the past decade, while the main studies showed a sub-
based on the debate between supporters of shareholder value and stantially neutral position in the early 2000s, albeit with a tendency to
those of stakeholder theory (Freeman, 1983). More specifically, the favor the superior performance of ESG investments. More recent
first theory refers to neoclassical economics and postulates that studies, on the contrary, recognize the positive and sometimes signifi-
the best contribution of a firm to social well-being lies in maximizing cant correlation between sustainable strategies and superior perfor-
its productivity and profitability (Levitt, 1958). In contrast to this mance compared with those resulting from traditional allocation
vision, the scientific literature of recent decades has favored stake- methodologies. In particular, de Haan et al. (2012) and Alessi
holder theory, which focuses on the reputational dimension of the et al. (2019) focus on environmental factors, Edmans (2012) identifies
firm and matches the complexity of the society in which it operates a strong positive relationship between social factors and the trend of
(Donaldson & Preston, 1995). stock prices, and Renders et al. (2010) find a positive relationship
Empirical studies have not thus far reached a consensus on this between corporate governance and corporate financial performance.
topic. Ballestero et al. (2015) trace the problem to possible biases in Further investigations into the positive links between corporate social
the data samples used in quantitative analyses because of issues such responsibility and financial performance have focused on specific eco-
as the length of the time series, sample representativeness, and nomic sectors (Brogi & Lagasio, 2019).
hypotheses underlying the analyses. The contradictions in the results
of previous studies lie in the nature of the factors on which they are
based. Indeed, empirical analyses of ESG mutual funds are character- 3 | DA T A SE T A N D M E TH OD O LO GY
ized by the employment of non-homogeneous asset pricing models,
which use different factors such as investors' cash flows This empirical analysis aims to verify the presence of positive devia-
(Bollen, 2007), fees (Gil-Bazo et al., 2010; Kreander et al., 2005), social tions between the performance of funds characterized by a higher
responsibility indices (Renneboog et al., 2008), and investment styles sustainability rating compared with funds with lower ratings. The
(Chen & Scholtens, 2018). To overcome these limitations, Ielasi and comparison of these two categories of funds requires the identifica-
Rossolini (2019) use traditional asset pricing models and factors, tion of two distinct datasets to limit the presence of bias in the data
applying them to different categories of sustainable funds to identify and maximize sample representativeness. Therefore, only funds with a
their peculiarities. specific set of characteristics were selected to narrow the field of
When examining issues related to sustainability and ESG con- observation and maximize the difference between the two groups in
cerns, the factors used are often qualitative in nature, and studies' terms of sustainability. In particular, we collected monthly returns
measurement methodologies frequently differ. However, recent from the beginning of October 2014 to the end of September 2019
empirical studies typically find a positive relationship between finan- of the oldest share class of equity open-end funds, domiciled in
cial performance and social performance. Friede et al. (2015) provide Europe, and denominated in Euro. The choice of a geographical selec-
a comprehensive review and systematization of the extensive scien- tion criterion is typical in this strand of the literature (Bauer
tific literature available, collecting and classifying more than 2200 et al., 2007; Cortez et al., 2009; Jin, 2018; Nofsinger & Varma, 2014)
empirical studies conducted since the 1970s, to identify the correla- and is also justified by the comparability of the institutional
tion between firms' attention to ESG criteria and their corporate framework.
financial performance. They find a non-negative relationship, although The data necessary for this analysis were downloaded from the
not always statistically significant, between attention to ESG criteria Morningstar Direct database. The selected funds were divided into
and corporate financial performance in approximately 90% of the two subsamples according to their Morningstar Sustainability Rating,
empirical studies examined. The robustness of this positive correlation a measure of the financially material ESG factors of the securities held
is further investigated by Brooks and Oikonomou (2018) and Busch by a fund (Morningstar Research, 2019). This rating is a normally dis-
and Friede (2018). tributed ordinal score and descriptive rank of a fund relative to the
The second branch of the scientific literature focuses on identify- fund's global category. Specifically, we selected only funds with either
ing positive deviations in the performance of sustainable investment a “high” or a “low” rating, which correspond to the top 10% and bot-
strategies compared with more traditional strategies. In this context, tom 10% of the distribution of the Morningstar Sustainability Ratings,
the fund management company plays a crucial role, with particular respectively.
reference to its ability to effectively implement asset allocation and The sample identified by this selection (634 funds) comprises
stock picking as well as its operational efficiency, which directly 442 mutual funds with high ESG ratings and 192 with low ESG rat-
impacts the cost structure and cost-effectiveness of the fund itself. ings. The average portion of assets under management scored by the
To assess the correlation between the integration of ESG criteria ESG rating is 93.92% for the former set of funds and 89.57% for
in the management of listed companies and the performance of their the latter. Table 1 reports the full descriptive statistics of the panel
securities, some conclusions can be drawn from two reviews of the of data.
 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2175 by University Of Embu, Wiley Online Library on [30/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1448 ABATE ET AL.

TABLE 1 Descriptive statistics of the sample

High ESG rating funds Low ESG rating funds Full sample
Expected return Mean 0.57% 0.45% 0.54%
Maximum 1.51% 2.31% 2.31%
Minimum 1.24% 0.55% 1.24%
Standard deviation Mean 3.93% 4.44% 4.07%
Maximum 10.15% 11.96% 11.96%
Minimum 0.83% 2.31% 0.83%
Skewness Mean 0.182 0.209 0.191
Maximum 1.143 1.456 1.456
Minimum 1.016 1.409 1.409
Kurtosis Mean 3.173 3.543 3.285
Maximum 5.981 9.721 9.721
Minimum 1.946 2.261 1.946
Downside deviation Mean 2.85% 3.23% 2.96%
Maximum 6.39% 7.19% 7.19%
Minimum 0.59% 1.52% 0.59%

The present study has been implemented using the DEA, a non- time window analyzed. Therefore, for each j-th fund we use the
parametric technique that attributes an efficiency score to each unit accumulation factor Uj (Basso & Funari, 2007), obtained by an
in the sample. DEA can be used in various contexts to evaluate the investment in the j-th fund in October 2014, net of front loads Fj,
performance of multi-input and multi-output systems. Its first applica- and kept until the end of September 2019:
tion in asset management was by Murthi et al. (1997) as a perfor-



mance measure to evaluate mutual funds, and it has become the Uj ¼ 1 – F j  1 þ Rj
reference basis for all subsequent studies in this field (Basso &
Funari, 2016). With: j = 1, 2, …, 634; Rj is the cumulative return, net of the ongo-
DEA has the following advantages over typical risk-adjusted per- ing charges.
formance measures: We have included the fees in the calculation of Uj, instead of
treating them separately, because of the higher computational effi-
• It does not require a benchmark to compare the performance of a ciency of this formulation. These cost measures are essential for a
fund, as the comparison ranks the performances of all the funds in thorough evaluation of each fund j by an investor. In fact, ongoing
the sample at the same time; charges represent the costs directly borne by the fund's assets and
• It is highly flexible and allows the incorporation of a large number thus are incurred indirectly by the investor. They provide an overall
of specific inputs and outputs; assessment of the main costs that weigh on the fund. Front-end loads,
• It allows the monitoring of the marginal contribution made by each instead, represent the costs directly borne by the investor. Mor-
input to overall performance. ningstar Direct provides the maximum percentage of fees reported in
the prospectus of the fund and consequently, to simulate the distribu-
After identifying the DEA efficiency measure for each fund, we can tor's ability to discount them, their amount has been reduced by 50%
evaluate the ranking of funds with high ESG ratings compared with in our analysis.
those with low ESG ratings. The DEA model used in this study fol- To evaluate the risk profile of the funds, we adopt statistical mea-
lows the technique proposed by Charnes et al. (1978), also known sures covering the higher moments of distributions, thus avoiding the
as CCR model, and aims to measure the net performance of each limiting assumption of normality. Therefore, the following three risk
fund based on the combination of the three risk measures deemed measures, calculated on the monthly returns during the five-year sam-
significant for an investor. More precisely, a measure of the net ple period, are used as inputs in the DEA for each fund j:
return reported by the funds within the considered time horizon is
used as the output. However, the choice of this measure requires • The standard deviation (Dj) as an absolute risk measure;
particular attention, as this DEA methodology can only be • The downside deviation (DDj) as an asymmetric risk measure,
implemented correctly with non-negative values. Consequently, it replacing the usual index of skewness, because it can take a nega-
was not possible to use the mean returns of the funds as the out- tive value (Gregoriou, 2007). It is calculated as follows (where rj,t is
put, because some of them reported negative values during the the return in month t and μj is the mean return):
 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2175 by University Of Embu, Wiley Online Library on [30/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABATE ET AL. 1449

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u 1 X
2 problem and this coincides with the value taken by the objective
n
DDj ¼ u r j,t  μj ð1Þ
un  1 function:
t t¼1
t:r j,t < μj
max uj Uj ð7Þ
ðuj ,vi,j Þ

• Kurtosis (Kj) as an absolute risk measure typical of non-normal

distributions. Subject to:

The following is the formulation of the fractional programming prob- v1,j Dj þ v2,j DDj þ v 3,j K j ¼ 1 ð8Þ
lem to be solved to calculate the DEA efficiency measure, in the CCR
model, for each fund j, with j = 1, 2,…, 634: uj Uj ≤ v 1,j Dj þ v2,j DDj þ v3,j K j j ¼ 1,2, …, 634 ð9Þ

uj Uj uj > 0 ð10Þ
max ð2Þ
ðuj ,vi,j Þ v1,j Dj þ v2,j DDj þ v3,j K j
vI,j > 0 i ¼ 1, 2,3 ð11Þ

where: uj is the weight of the output; vi,j (with i = 1, 2, 3) is the weight

of each input. Subject to: As for the constraint (3), also constraint (9) must hold for the full sam-
ple of funds at the same time. By solving this optimization problem
uj Uj for each fund j, we can calculate its efficiency measure, which can
≤ j ¼ 1, 2,…, 634 ð3Þ
v1,j Dj þ v2,j DDj þ v 3,j K j
take a value in the range [0, 1]. In particular, fund j reaches maximum
efficiency with respect to the sample if its objective function (7) takes
uj > 0 ð4Þ
a value equal to 1; otherwise, the fund is considered to be inefficient
because its input–output combination is dominated by that of at least
v i,j > 0 i ¼ 1, 2,…, 5 ð5Þ
one other fund of the same sample.
While the performance and risk measures employed in this analysis
Since the optimization is carried out for each j-th mutual fund, the are typical of the asset management industry, the inclusion of cost mea-
optimal value of the weights will be different, in general, for each sures in the calculation of Uj deserves further clarification. A fund seeking
one. Note that inequality (3) must hold for all the funds subject to a high ESG rating may be subject to a more expensive cost structure
the DEA at the same time; in other words, no fund can reach a level than that of a conventional fund. This potential cost increase arises
of efficiency above the maximum value of 1 by making use of the because, in addition to the usual analysis based exclusively on financial
weights calculated for another fund from the same set. Therefore, criteria, sustainable investments must take into account ESG factors, for
the DEA efficiency measure is a relative metric and varies according which it is necessary to resort to control or certification bodies represen-
to the sample, providing a ranking valid only within the sample ted by an ethics committee and ESG advisors/rating agencies. Conse-
itself. quently, this greater complexity should generally translate into an
To simplify the resolution of this fractional programming problem, increase in the operating costs incurred by the management company
it is useful to transform it into an equivalent linear programming prob- and, therefore, in the fees charged to the fund and the investors.
lem (Cooper et al., 2007). Indeed, the denominator in inequality (3),
under the constraint of non-negativity of the weights (5) and under
the assumption that the input values are always non-negative, is non- 4 | EMP I RICAL A NALY SI S
negative for each j. Therefore, it is possible to multiply both sides of
inequality (3) by its denominator without having to reverse the sign of The objective of the DEA carried out on this panel of data is to verify
the latter. After making the appropriate simplifications, the following whether funds with high ESG-rated portfolios outperform funds that
inequality is obtained for each fund j: hold portfolios of low ESG-rated securities, thus highlighting a poten-
tial financial incentive in truly sustainable investments. Table 2 reports
uj Uj ≤ v1,j Dj þ v2,j DDj þ v 3,j K j ð6Þ the main statistics for the DEA efficiency measure for the two
datasets and full sample.
At this point, since the output-oriented logic of DEA seeks to maxi- The DEA scores between the two types of funds differ. The mean
mize the output with the same amount of inputs, to transform the value achieved by funds with high ESG ratings is 0.661 compared with
fractional form of Equation (2) into a more tractable linear form it is 0.562 for low ESG-rated funds. This evidence could suggest a possible
sufficient to set the denominator of the objective function (2) equal preference for funds with high ESG ratings. However, to verify
to 1 and add inequality (6) to the constraints of the optimal prob- whether this divergence between the two measures is statistically sig-
lem. Therefore, the relative efficiency measure for each fund j is nificant, as a first step we carry out a unilateral hypothesis t-test. The
the solution of the following equivalent linear programming hypothesis system is as follows:
 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2175 by University Of Embu, Wiley Online Library on [30/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1450 ABATE ET AL.

TABLE 2 DEA efficiency measure

High ESG funds Low ESG funds All funds
Sample size 442 192 634
Number of efficient funds 4 0 4
Mean DEA efficiency measure 0.661 0.562 0.631
Maximum DEA efficiency measure 1 0.884 1
Minimum DEA efficiency measure 0.193 0.189 0.189
Median DEA efficiency measure 0.667 0.584 0.641

F I G U R E 1 Cumulative distribution
of the DEA efficiency measures

H0 : μh  μl ¼ 0 • H0: the DEA efficiency measures of high ESG-rated funds and low
ð12Þ
H1 : μh  μl > 0 ESG-rated funds come from populations with the same
distribution;
where μh: mean of the DEA efficiency measures of high ESG-rated • H1: the empirical cumulative distribution function of the DEA effi-
funds; μl: mean of the DEA efficiency measures of low ESG-rated ciency measures of high ESG-rated funds lies below the empirical
funds.The t-statistic of this test takes a value of 8.327, rejecting the cumulative distribution function of the DEA efficiency measures of
null hypothesis with significance at the 1% level and, therefore, indi- low ESG-rated funds.
cating a positive difference in favor of funds with high ESG ratings in
the sample analyzed. This initial test, anyway, requires further investi- As it could be expected from the cumulative distribution functions
gation before we can draw any conclusion. shown in Figure 1, the test statistic takes a value of 0.359 and there-
Figure 1 shows the cumulative distribution functions of the DEA fore the Kolmogorov–Smirnov test rejects the null hypothesis with
efficiency measures. Almost 65% of funds with high ESG ratings significance at the 1% level, providing an additional evidence of the
obtain a DEA efficiency measure of at least 0.63, while only 29.7% of superior efficiency of high-ESG rated funds.
funds with low ESG ratings reach a similar score. If we subdivide the We now analyze each variable used in the DEA model to identify
distribution into 20 quantiles, the highest frequency for funds with those that contributed the most to the difference between the two
high ESG ratings is 20.81% and this is associated with DEA values subsamples. Figure 2 shows the distributions of the net accumulation
between 0.65 and 0.70. With regard to funds with low ESG ratings, factors for both sets of funds. More than 40% of funds with high ESG
on the contrary, the highest frequency is 19.27% and this is associ- ratings generate a cumulative performance 1.25–1.45 times the initial
ated with DEA values between 0.60 and 0.65. unitary investment in the time horizon analyzed. This figure also high-
In addition to the t-test, following Banker and Natarajan (2011) lights the significantly larger percentage of funds with low ESG ratings
we have employed a two-sample Kolmogorov–Smirnov one-sided (17.19%) that generate a negative cumulative return if compared to
hypothesis test, which evaluates the vertical distance between the the funds with high ESG ratings (3.85%).
cumulative distribution functions of the two sets of DEA efficiency Figure 3 shows the distribution of the relative frequencies of the
measures. The hypothesis system is as follows: monthly standard deviations of returns, which is useful for assessing
15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2175 by University Of Embu, Wiley Online Library on [30/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1451

F I G U R E 2 Frequency distribution

F I G U R E 3 Frequency distribution

F I G U R E 4 Frequency distribution
of the net accumulation factors

of standard deviations

of ongoing charges
ABATE ET AL.
 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2175 by University Of Embu, Wiley Online Library on [30/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1452 ABATE ET AL.

TABLE 3 Statistical significance of the distance between each pair of variables

High ESG rating funds Low ESG rating funds t-test statistic K.-S. test statistic
***
Accumulation factor (net) 1.361 1.268 3.672 0.232***
Standard deviation 3.93% 4.44% 4.527*** 0.204***
Downside deviation 2.85% 3.23% 5.022 ***
0.255***
Kurtosis 3.173 3.543 5.106*** 0.240***
Ongoing charges 1.53% 1.92% 6.170 ***
0.247***
Reduced front-end loads 1.24% 1.31% 0.749 0.117**

Note: ***, **, and * represent statistical significance levels of 1%, 5%, and 10%, respectively.

TABLE 4 Summary statistics of the DEA on homogeneous segments of the equity asset class

Europe equity Global equity Other equity

High Low t-test K.-S. test High Low t-test K.-S. test High Low t-test K.-S. test
ESG ESG statistic statistic ESG ESG statistic statistic ESG ESG statistic statistic
Sample size 196 83 108 72 138 37
Efficient funds 2 0 3 0 3 0
DEA measure 0.748 0.614 6.670*** 0.488*** 0.748 0.685 3.531*** 0.319*** 0.647 0.542 3.618*** 0.302***
*** *** *** ***
Accumulation factor (net) 1.319 1.209 3.138 0.324 1.385 1.280 3.446 0.370 1.403 1.374 0.348 0.164
Standard deviation 4.09% 4.60% 4.618*** 0.362*** 3.63% 3.87% 2.172** 0.264*** 4.24% 5.18% 2.593*** 0.254**
Downside deviation 2.97% 3.38% 4.979*** 0.404*** 2.72% 2.84% 1.523* 0.190** 3.02% 3.65% 3.012*** 0.333***
Kurtosis 2.90 3.24 4.618*** 0.372*** 3.49 3.72 2.064** 0.167* 3.51 3.87 2.633*** 0.267**
Ongoing charges 1.57% 1.85% 4.312 ***
0.260 ***
1.53% 1.93% 3.608 ***
0.306 ***
1.64% 2.05% 2.986*** 0.286***
Reduced front-end loads 1.26% 1.28% 0.192 0.088 1.42% 1.44% 0.107 0.106 1.10% 1.14% 0.233 0.160

Note: ***, **, and *represent statistical significance levels of 1%, 5%, and 10%, respectively.

fund risk. Overall, funds with high ESG ratings tend to be less subject These tests confirm the presence of a statistically significant gap
to absolute risk than funds with low ESG ratings. To clarify this intui- in favor of funds with high ESG ratings for all the variables considered,
tion, note that almost 62% of funds with high ESG ratings have a with the partial exception of front-end loads. Funds that apply ESG
monthly standard deviation below 4% compared with slightly less criteria to stock selection achieve considerably superior risk-adjusted
than 47% of funds with low ESG ratings. performances than funds with low ESG ratings, thanks not only to
With regard to the cost measures, the disparity between the two their higher returns, but also to their better risk control, including
sets of funds is derived from their ongoing charges. Figure 4 shows higher moments. The presented evidence, added to that obtained
the distribution of the relative frequencies of this variable. Most of from the DEA model, suggests a positive correlation between social
the funds analyzed are subject to ongoing charges between 1.5% and performance and financial performance.
2.5% of the assets. However, lower levels of ongoing charges have Moreover, the ongoing charges of high ESG-rated funds are sig-
significantly higher relative frequencies for funds with high ESG rat- nificantly lower than those of their counterparts, a feature unreported
ings. By contrast, about 46% of funds with low ESG ratings incur an by the scientific literature. Indeed, the first research conducted in this
annual cost of more than 2% of their assets under management. area in the early 2000s did not find a positive correlation between the
Unilateral hypothesis tests were conducted for each pair of vari- management fees charged by a fund and the number of screening
ables to understand whether the difference in the DEA efficiency criteria applied (Young & Proffitt, 2003). Moreover, the limited num-
measure is attributable to statistically significant deviations between ber of studies available have not measured substantial differences in
the input and output of the two sets of funds. Table 3 shows the terms of fees between sustainable and conventional funds (Gil-Bazo
mean values of the variables and the test statistics of the unilateral et al., 2010; Kreander et al., 2005). The rationale for this discrepancy
hypothesis t-tests and Kolmogorov–Smirnov tests carried out on each with previous research may be due not only to the different samples
pair. The alternative hypothesis is that the tested variable of funds employed, but also to the fact that our study focuses on the two
with high ESG ratings has a more favorable mean (t-test) or cumula- extremes of the ESG spectrum, that is, the best and worst European
tive distribution function (K.-S. test) to an investor compared with funds in terms of ESG rating, thereby avoiding funds with mixed
funds with low ESG ratings. approaches to sustainable investments.
 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2175 by University Of Embu, Wiley Online Library on [30/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABATE ET AL. 1453

TABLE 5 Summary statistics of the DEA on 1000 simulated pairs of sets

t-test Kolmogorov–Smirnov test

H1 accepted (number H1 accepted (number

of simulations) Mean test statistic of simulations) Mean test statistic
***
DEA efficiency measure 999 5.064 998 0.375***
Accumulation factor (net) 758 2.453* 935 0.257**
Standard deviation 943 3.090 **
921 0.245**
Downside deviation 970 3.412*** 984 0.286***
Kurtosis 981 3.375 ***
962 0.267***
Ongoing charges 988 3.881*** 975 0.275***
Reduced front-end loads 112 0.450 295 0.131

Note: ***, **, and * represent statistical significance levels of 1%, 5%, and 10%, respectively.

TABLE 6 Sharpe ratios

Sharpe ratio

High ESG Low ESG t-test statistic K.-S. test statistic

***
All funds 0.157 0.116 5.739 0.245***
***
Europe 0.143 0.101 3.963 0.344***
Global 0.173 0.133 3.970*** 0.394***
***
Other 0.164 0.118 2.463 0.191

Note: ***, **, and * represent statistical significance levels of 1%, 5%, and 10%, respectively.

The only variable indistinguishable between the two sets is front- provides a robustness test of the empirical results. Table 5 reports, for
end loads. Given that their amount is more linked to the distribution each variable, the number of simulations in which H1 (i.e., the superior
policies and intermediaries involved than the nature of the funds, this efficiency of high ESG-rated funds) is accepted by a t-test and a
result is expected. Kolmogorov–Smirnov test, both with a statistical significance of at
To verify that these results are not strictly dependent on the sam- least 5%.
ple employed, we applied robustness tests. In particular, we aimed to A further check is carried out to address the possible presence of
address two potential concerns: the exposure of each set of funds model risk in the DEA implemented in this study. In other words, we
to distinct segments within the equity asset class and the different measure the efficiency of the two samples of funds resorting to the
sizes of the two sets. Sharpe ratio (Sharpe, 1966), a different and more traditional risk-
The first potential bias was evaluated by subdividing the sets of adjusted performance measure. The index FTSE EUR EuroDep 1 Mon
high ESG-rated and low ESG-rated funds into three homogeneous EUR is employed as the proxy of the risk-free rate and fund returns
subsets, so that the funds of each pair of subsets belong to a segment are net of ongoing charges. The full sample has also been divided into
of the broader Equity asset class. In more detail, we identified three the same homogeneous geographical segments used in Table 4.
segments based on the Morningstar Global Category of the funds: Table 6 summarizes the results of this analysis, confirming the out-
Europe Equity, Global Equity, and Other Equity. Table 4 reports the come of the DEA and therefore the superior performance of mutual
means of the DEA efficiency measures and of each variable involved, funds with high ESG rating.
and both the t-test and the Kolmogorov–Smirnov test show statisti- The empirical evidence presented in these analyses agrees with
cally significant differences between each pair. This confirms the studies of the US stock market (Auer & Schuhmacher, 2016; Eccles
superior efficiency of funds with high ESG ratings, with limited differ- et al., 2014; Khan et al., 2016) and the stocks of European banks
ences among the three sectors of the Equity asset class.  s et al., 2019), which have reached similar results, con-
(Miralles-Quiro
For the second robustness check, we tested whether the size of cluding that the best securities from a sustainability point of view
each set of funds could affect the DEA results. This analysis was run achieve significantly higher market risk-adjusted performance than
by applying the DEA to 1000 simulated pairs of sets, each composed companies that have not adopted sustainable practices. Hence, the
of 100 funds (with a high and a low ESG rating) randomly sampled empirical analysis presented here shows that the link between finance
from the original data. With this methodology, the size of the sets is and sustainability can generate profitable synergies. This result con-
always the same, and the use of a large sample simulation technique curs with the empirical study by Dur
an-Santomil et al. (2019), focusing
 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2175 by University Of Embu, Wiley Online Library on [30/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1454 ABATE ET AL.

on European sustainable funds, whereas it contrasts with that of actual ESG investment policy of funds. While this is the result of a
Dolvin et al. (2019), which takes US funds into account. This discrep- detailed company-level analysis, it does not provide any guarantee
ancy should be the subject of further research, as both these studies, that the portfolio manager complied with ESG criteria before the rat-
like the present one, used samples of funds with Morningstar Sustain- ing process started, or will in the future. However, unless the manage-
ability Ratings. ment policy is highly active, which implies a substantial turnover, the
sustainability of the fund should be stable. Another suboptimal aspect
of our sample is its relatively short length, 60 months, because of the
5 | C O N CL U S I O N S trade-off with sample size: a longer time series would have meant a
significant reduction in the number of available funds. Future research
The results of our empirical analysis agree with the branch of the litera- would benefit from the increasing number of funds with ESG ratings
ture supporting the existence of a positive relationship between ESG to overcome this limitation.
performance and financial performance. In particular, our study provides
an innovative point of view, because we selected funds at the two RE FE RE NCE S
extremes of the spectrum of ESG ratings, thereby excluding funds with Alessi, L., Ossola, E., & Panzica, R. (2019). The Greenium matters: Evidence
a blended approach to sustainable investment. Indeed, the sample of on the pricing of climate risks. Publications Office of the European
Union. https://doi.org/10.2760/403111
funds with high Morningstar Sustainability Ratings achieved significantly
Amel-Zadeh, A., & Serafeim, G. (2018). Why and how investors use ESG
higher average DEA efficiency than the sample of funds with low rat- information: Evidence from a global survey. Financial Analysts Journal,
ings. Our results are robust to different sectors of the equity asset class 74(3), 87–103. https://doi.org/10.2469/faj.v74.n3.2
and were verified through the resampling of 1000 paired sets of funds. Auer, B., & Schuhmacher, F. (2016). Do socially (ir)responsible investments
pay? New evidence from international ESG data. The Quarterly Review
Ultimately, investors who decide to allocate their resources in
of Economics and Finance, 59(C), 51–62. https://doi.org/10.1016/j.
equity funds of European asset managers with a preference for ESG qref.2015.07.002
criteria would obtain superior financial efficiency than investing in Ballestero, E., Pérez-Gladish, B., & Garcia-Bernabeu, A. (2015). Socially
similar funds with less sustainable investment policies. Indeed, invest- responsible investment. A multi-criteria decision making approach.
Springer. https://doi.org/10.1007/978-3-319-11836-9
ment policies based on ESG criteria seem to enjoy a competitive
Banker, R. J., & Natarajan, R. (2011). Statistical tests based on DEA effi-
advantage because of their inclusion of non-financial data. Moreover,
ciency scores. In W. W. Cooper, L. M. Seiford, & J. Zhu (Eds.), Hand-
our findings show that funds with high ESG ratings are subject to book on data envelopment analysis (2nd ed.). Springer. https://doi.org/
lower ongoing charges despite their more complex screening activities 10.1007/978-1-4419-6151-8
and associated operating costs. Basso, A., & Funari, S. (2007). DEA models for ethical and non ethical mutual
funds. Mathematical Methods in Economics and Finance, 2, 21–40.
Attention to ESG policies should also be a primary concern for
Basso, A., & Funari, S. (2016). DEA performance assessment of mutual
firms. The results of our analysis show the relevance of ESG criteria funds. In J. Zhu (Ed.), Data envelopment analysis. A handbook of empiri-
for mutual fund managers, especially those providing superior risk- cal studies and applications. Springer. https://doi.org/10.1007/978-1-
adjusted performance to investors. As a consequence, firms that 4899-7684-0_8
Bauer, R., Derwall, J., & Otten, O. (2007). The ethical mutual fund perfor-
employ sustainability strategies should pay attention to their corpo-
mance debate: New evidence from Canada. Journal of Business Ethics,
rate social responsibility reports and solicit an ESG rating to be 70(2), 111–124. https://doi.org/10.1007/s10551-006-9099-0
targeted by fund managers and therefore increase shareholder value. Becchetti, L., Ciciretti, R., Dalò, A., & Herzel, S. (2015). Socially responsible
Further, our findings provide relevant information for and conventional investment funds: Performance comparison and the
global financial crisis. Applied Economics, 47(25), 2541–2562. https://
policymakers. Current classifications of sustainable funds are mainly
doi.org/10.1080/00036846.2014.1000517
based on qualitative criteria, while objective ones are often difficult to
Bollen, N. (2007). Mutual fund attributes and investor behavior. Journal of
identify. On the contrary, we showed that ESG ratings can be an Financial and Quantitative Analysis, 42(3), 683–708. https://doi.org/10.
effective informative instrument and, therefore, their role should be 1017/S0022109000004142
taken into account by supervisory authorities, for example, by requir- Brogi, M., & Lagasio, V. (2019). Environmental, social, and governance and
company profitability: Are financial intermediaries different? Corporate
ing their inclusion in the key investor information document of mutual
Social Responsibility and Environmental Management, 26(3), 576–587.
funds. In the European Union, ESMA (2020) acknowledged the rele- https://doi.org/10.1002/csr.1704
vance of ESG ratings in the risk assessment of investments, but our Brooks, C., & Oikonomou, I. (2018). The effects of environmental, social
study provides further evidence in favor of their wider employment and governance disclosures and performance on firm value: A review
of the literature in accounting and finance. The British Accounting
(e.g., in the disclosure of sustainable funds' investment policies). The
Review, 50(1), 1–15. https://doi.org/10.1016/j.bar.2017.11.005
absence of a common valuation metric, however, poses problems of Busch, T., & Friede, G. (2018). The robustness of the corporate social and
comparability among ratings, which should be the subject of further financial performance relation: A second-order meta-analysis. Corpo-
scrutiny by scholars and regulators alike. rate Social Responsibility and Environmental Management, 25(4), 583–
608. https://doi.org/10.1002/csr.1480
Our research is subject to limitations related to the data
Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efficiency
employed, which should be taken into consideration. One limitation is of decision making units. European Journal of Operational Research, 2
the use of the Morningstar Sustainability Rating as a proxy for the (6), 429–444. https://doi.org/10.1016/0377-2217(78)90138-8
 15353966, 2021, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/csr.2175 by University Of Embu, Wiley Online Library on [30/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABATE ET AL. 1455

Chen, X., & Scholtens, B. (2018). The urge to act: A comparison of active Jin, I. (2018). Is ESG a systematic risk factor for US equity mutual funds?
and passive socially responsible investment funds in the United States. Journal of Sustainable Finance & Investment, 8(1), 72–93. https://doi.
Corporate Social Responsibility and Environmental Management, 25(6), org/10.1080/20430795.2017.1395251
1154–1173. https://doi.org/10.1002/csr.1529 Khan, M., Serafeim, G., & Yoon, A. (2016). Corporate sustainability: First
Clark, G. L., Feiner, A., & Viehs, M. (2015). From the stockholder to the stake- evidence on materiality. The Accounting Review, 91(6), 1697–1724.
holder. How sustainability can drive financial outperformance. [Working https://doi.org/10.2308/accr-51383
Paper] Oxford-London: University of Oxford/Arabesque Partners. Kreander, N., Gray, R. H., Power, D. M., & Sinclair, C. D. (2005). Evaluating
Cooper, W. W., Seiford, L. M., & Tone, K. (2007). Data envelopment analysis. the performance of ethical and non-ethical funds: A matched pair anal-
A comprehensive text with models, applications, references and DEA-solver ysis. Journal of Business Finance & Accounting, 32(7–8), 1465–1493.
software (2nd ed.). Springer. https://doi.org/10.1007/978-0-387- https://doi.org/10.1111/j.0306-686X.2005.00636.x
45283-8 Levitt, T. (1958). The dangers of social responsibility. Harvard Business
Cortez, M. C., Silva, F., & Areal, N. (2009). The performance of European Review, 36(5), 41–50.
socially responsible funds. Journal of Business Ethics, 87(4), 573–588. Miralles-Quiro  s, M. M., Miralles-Quiro
 s, J. L., & Redondo-Hernandez, J.
https://doi.org/10.1007/s10551-008-9959-x (2019). The impact of environmental, social, and governance perfor-
de Haan, M., Dam, L., & Scholtens, B. (2012). The drivers of the relation- mance on stock prices: Evidence from the banking industry. Corporate
ship between corporate environmental performance and stock market Social Responsibility and Environmental Management, 26(6), 1446–
returns. Journal of Sustainable Financial Investment, 2(3–4), 338–375. 1456. https://doi.org/10.1002/csr.1759
https://doi.org/10.1080/20430795.2012.738601 Morningstar Research. (2019). Morningstar sustainability rating methodol-
Dolvin, S., Fulkerson, J., & Krukover, A. (2019). Do “good guys” finish last? ogy. Morningstar.
The relationship between Morningstar sustainability ratings and Murthi, B. P. S., Choi, Y. K., & Desai, P. (1997). Efficiency of mutual funds
mutual fund performance. The Journal of Investing, 28(2), 77–91. and portfolio performance measurement: A non-parametric approach.
https://doi.org/10.3905/joi.2019.28.2.077 European Journal of Operational Research, 98(2), 408–418. https://doi.
Donaldson, T., & Preston, L. E. (1995). The stakeholder theory of the cor- org/10.1016/S0377-2217(96)00356-6
poration: Concepts, evidence and implications. The Academy of Man- Nofsinger, J., & Varma, A. (2014). Socially responsible funds and market
agement Review, 20(1), 65–91. https://doi.org/10.5465/amr.1995. crisis. Journal of Banking and Finance, 48, 180–193. https://doi.org/10.
9503271992 1016/j.jbankfin.2013.12.016
Duran-Santomil, P., Otero-Gonzalez, L., Correia-Domingues, R. H., & Renders, A., Gaeremynck, A., & Sercu, P. (2010). Corporate-governance
Reboredo, J. C. (2019). Does sustainability score impact mutual fund ratings and company performance: A cross-European study. Corporate
performance? Sustainability, 11(10), 2972. https://doi.org/10.3390/ Governance: An International Review, 18(2), 87–106. https://doi.org/
su11102972 10.1111/j.1467-8683.2010.00791.x
Eccles, R. G., Ioannou, I., & Serafeim, G. (2014). The impact of corporate sus- Renneboog, L., Ter Horst, J., & Zhang, C. (2008). The price of ethics and
tainability on organizational processes and performance. Management stakeholder governance: The performance of socially responsible
Science, 60(11), 2835–2857. https://doi.org/10.1287/mnsc.2014.1984 mutual funds. Journal of Corporate Finance, 14(3), 302–322. https://
Edmans, A. (2012). The link between job satisfaction and firm value, with doi.org/10.1016/j.jcorpfin.2008.03.009
implications for corporate social responsibility. Academy of Management Sharpe, W. F. (1966). Mutual fund performance. Journal of Business, 39
Perspectives, 26(4), 1–19. https://doi.org/10.5465/amp.2012.0046 (S1), 119–138. https://doi.org/10.1086/294846
Engle, R. F., III, Giglio, S., Kelly, B., Lee, H., & Stroebel, J. (2020). Hedging United Nations Environment Programme Finance Initiative, & Mercer
climate change news. The Review of Financial Studies, 33(3), 1184– (2007). Demystifying responsible investment performance. In A review
1216. https://doi.org/10.1093/rfs/hhz072 of key academic and broker research on ESG factors. UNEP FI/Mercer.
ESMA. (2020). Strategy on sustainable finance, ESMA22-105-1052. van Duuren, E., Plantinga, A., & Scholtens, B. (2016). ESG integration and
European Securities and Markets Authority. the investment management process: Fundamental investing
Freeman, R. E. (1983). Strategic management: A stakeholder approach. reinvented. Journal of Business Ethics, 138(3), 525–533. https://doi.
Advances in Strategic Management, 1(1), 31–60. org/10.1007/s10551-015-2610-8
Friede, G., Busch, T., & Bassen, A. (2015). ESG and financial performance. Young, K. & Proffitt, D. (2003). Socially responsible mutual funds: Recent
Aggregated evidence from more than 2000 empirical studies. Journal performance and other issues relating to portfolio choice. [Working
of Sustainable Finance & Investment, 5(4), 210–233. https://doi.org/10. Paper] Phoenix: College of Business and Professional Studies - Grand
1080/20430795.2015.1118917 Canyon University.
Gil-Bazo, J., Ruiz-Verdú, P., & Santos, A. A. P. (2010). The performance of
socially responsible mutual funds: The role of fees and management
companies. Journal of Business Ethics, 94(2), 243–263. https://doi.org/
10.1007/s10551-009-0260-4
Gregoriou, G. N. (2007). Efficiency of US mutual funds using data envelop- How to cite this article: Abate, G., Basile, I., & Ferrari, P.
ment analysis. In G. N. Gregoriou (Ed.), Performance of mutual funds. An (2021). The level of sustainability and mutual fund
international perspective. Palgrave Macmillan. https://doi.org/10.1057/ performance in Europe: An empirical analysis using ESG
9780230626492_8
ratings. Corporate Social Responsibility and Environmental
Ielasi, F., & Rossolini, M. (2019). Responsible or thematic? The true nature
of sustainability-themed mutual funds. Sustainability, 11(12), 3304.
Management, 28(5), 1446–1455. https://doi.org/10.1002/
https://doi.org/10.3390/su11123304 csr.2175