Bekaert Et Al. 2022
Bekaert Et Al. 2022
Bekaert Et Al. 2022
SDG’s1
SSRN Working Paper
Geert Bekaert,3 Richard Rothenberg,2 Miquel Noguer2
November 2022
Abstract
1
We thank the Office of Investment Management at the United Nations Joint Staff Pension Fund for
initiating the original idea of this research and their work on testing the alpha assumption of the active
ESG momentum portfolio and Madelyn Antoncic for early contributions to this research. The views
expressed in this paper are those of the authors and do not necessarily reflect the views of the United
Nations.
2
Global AI Corp.
3
Columbia Business School
“ESG” investing has become all the rage. Trillions of dollars are now invested taking
environmental (e.g., carbon emissions), social (e.g., fair labor practices) and
governance (e.g., internal corruption) (ESG) issues into account. ESG investing is
now the most popular form of “sustainable” investing, growing simultaneously with
companies focusing more on their long-term sustainability and the needs of all
stakeholders.
An important question is to assess how many asset managers are truly “walking the
talk.” A 2019 survey of RBC Global Asset Management found that less than 25% of
asset managers and asset owners “significantly” use ESG principles as part of their
investment approach and decision making (RBC Global Assets, 2019).
This may not be surprising as ESG issues may well conflict with the fiduciary duties
of asset owners and managers to act in the best interest of their beneficiaries. A
recent Department of Labor (DOL) proposal regarding the use of ESG risk factors in
Employee Retirement Incomes Security Act of 1974 (ERISA) accounts is consistent
with this view of a conflict. The new proposal states that “private employer-
sponsored retirement plans are not vehicles for furthering social goals or policy
objectives that are not in the financial interest of the plan. Rather, ERISA plans
should be managed with unwavering focus on a single, very important social goal:
providing for the retirement security of American workers.” (Department of Labor,
2020).
There is clearly also a “need for a systematic way to measure and assess how asset
managers execute ESG” (Kim and Yoon, 2020). However, there is a lack of generally
accepted agreed-upon standards and reporting requirements. While corporations now
largely self-report some ESG data, “the practice has been widely criticized for lacking
the rigor of traditional financial reporting,” which results in significant “green-
washing” and data biases. Thus, investors lack high-quality, firm-level ESG data, to
serve as key inputs in assessing, managing, and monitoring the ESG risks and
opportunities that a company faces” (Antoncic, 2019a). In fact, 63% of hedge funds
polled by KPMG responded to a recent survey that ESG investing is “hampered by
the lack of robust reliable data.” (KPMG, 2020).
Due to the lack of agreed ESG standards, major discrepancies exist across vendors
who rate, rank and provide company ESG scores. In fact, comparing a company’s
The U.N. Sustainable Development Goals (SDGs) are a much broader set of
sustainability issues than traditional ESG issues and focus on “good health and well-
being, the elimination of poverty, zero hunger, quality education, clean water and
sanitation, reduced inequity,” as well as the environment and other issues
encapsulated in ESGs. Most importantly, the SDGs call for “leaving no one behind.”
The SDGs have more factors and address the full spectrum of global macro systemic issues
that matter to all stakeholders, all businesses and all countries. The SDGs “are a
universal call to action” established in 2015 by 193 countries “to end poverty, protect
the planet and ensure that all people enjoy peace and prosperity by the year 2030.”
(United Nations (2020).
In this article, we explore the possibility of creating an active portfolio that achieves the
goals associated with ESG investing but still generates alpha, consistent with fiduciary
duties. In addition, we measure the SDG impact of the resulting active portfolio relative
to the benchmark. Specifically, we consider the US MSCI Index as the benchmark to
beat. Among the roughly 600 stocks in the index, we create an active portfolio of about
50 stocks using the MSCI ESG ratings which show positive ESG momentum, to
measure ESG performance and track its performance relative to the index. We find
that the portfolio significantly outperforms the index when relative momentum is used
and this outperformance persists when controlling for the Fama French three- (Fama
3
The remainder of the paper is organized as follows. Section II describes the ESG data,
the methodology to create an active portfolio, and contains detailed portfolio results.
Section III describes the SDG scores in some detail, and characterizes the SDG
footprint of the selected portfolio. Section IV concludes.
In this section, we describe the ESG data, the active portfolio construction, and its
performance.
ESG Database
MSCI ESG ratings are widely used by the investment community as a proxy for ESG
performance.2 The MSCI coverage universe is based on major MSCI indices (such as
the MSCI World Index), which include the world’s largest and most liquid stocks. For
a detailed description of the MSCI’s methodology, see MSCI (2019) and Serafeim
(2020); we provide a short summary here.
MSCI attempts to quantify the risk and opportunity exposure of each company on 37
so-called “Key Issues.” These issues are divided into three pillars (environmental,
social and governance) which correspond to one of ten macro themes identified by
MSCI as a concern to investors, inter alia; climate change, pollution and waste,
1
Two of the authors work for Global AI, whereas the first is an external consultant to the company.
2Alternative ratings are available, see Walter (2019) for a survey, and a discussion of some
conceptual and practical issues plaguing such ratings. Berg, Koelbel and Rigobon (2019)
quantitatively studies the differences across ratings from different rating agencies.
4
MSCI aggregates the key issue data to an overall score where each key issue is weighted
according to its assessed materiality in each industry. Given that ESG issues tend to
vary systematically across industries, MSCI calculates an industry-adjusted score so that
the actual ratings are industry specific and comparisons across industries are not
meaningful.
ESG Investing
Incorporating ESG into the investment process is not without challenges. If firms with
high ESG scores manage to lower their cost of capital by their ESG actions and/or
increase their future cash flows by avoiding certain risks, all else equal, firms with good
ESG performance would be valued more highly than similar firms with less exemplary
ESG performance.1 If a lower cost of capital is the source of the valuation premium, it
should be associated with lower returns going forward. Clearly, this might clash with
the fiduciary duty of some institutional investors.
Several research papers written by MSCI show evidence that MSCI ESG rating changes
(“ESG momentum”) may be a useful financial indicator (Giese et al., 2019; Giese and
Nagi, 2018). Companies with higher ESG ratings, on average, experienced fewer stock-
1
Recent research from AMUNDI, a large French asset management company (Bennani, Le
Guenedal, Lepetit, Ly, Mortier, Roncalli, Sekine, 2019) suggests that ESG could become a risk factor
itself, if most investors use ESG scores in their decision to over- or under- weight a company’s stock
in their portfolio.
5
To test the potential alpha due to the change in a US stock’s ESG score, we construct
two sector neutral portfolios – one on the basis of the relative percentage change in
the industry-adjusted ESG score, and the other on the basis of the absolute change in
industry-adjusted ESG scores. Using the 11 GICS (Global Industry Classification
Standard) sectors stocks in each sector are ranked, at the end of each year, based on
their absolute and relative ESG momentum. The 10% highest ranking stocks in each
of the 11 GICS sectors based on their absolute and relative ESG momentum are then
selected for inclusion in the portfolios. These stocks are held for a full year, after
which the portfolios are rebalanced. The stocks within each industry, and the industry
-portfolios themselves are market value-weighted. Appendix A describes the portfolio
construction in more detail. 1
The portfolios performance over the past 6 years is then analyzed against its relevant
benchmark, using the MSCI US index. We chose the MSCI US index as a benchmark
1
In this paper we use the United Nations joint Staff Pension Fund portfolio, which excludes the
tobacco and weapons industries. ESG investing raises an important question of whether there is a
“cost to being good, ” which is particularly vexing because of the poor quality of the data and the
fact that ESG funds frequently exclude companies based on various criteria, which can create
conflicts with fiduciary duty. SDG investing does not seek to exclude any company but instead
measures their impact to society across a variety of angles. This means that while the ESG approach
may well reduce investment flows to certain sectors, an SDG-focused approach can be used as an
objective investment tool for the assessment of non-financial risks and can help identify positive and
negative spillover effects that go far beyond the narrow ESG lens. The fact that SDGs are applicable
to investments at the corporate, infrastructure, and sovereign levels, makes it a powerful alternative
to traditional ESG investing.
6
Our analysis tracks both the daily and monthly returns of the active ESG portfolio and
the MSCI US index for the selected sample January 2013 - December 2018. Figure 1
plots the cumulative return performance over the sample period, showing the relative
ESG momentum portfolio having the best performance followed by the absolute ESG
momentum portfolio, and the benchmark index performing worst. For the purposes of
this analysis, the most important test is whether the portfolio provides alpha with
respect to the relevant benchmark. Table 1 (Panel A) reports the constructed portfolio
alphas and the betas with respect to the index return (Market Model). The portfolio
alphas are also shown relative to the Fama-French three- and five-factor models. In
addition, we report the factor exposures to verify whether the ESG portfolios show
particular tilts relative to existing factors. The Fama-French (1993) three-factor model
adds two portfolios to the market model: the Small Minus Big (SMB) portfolio,
representing the return difference between an index of small versus an index of large
capitalization firms, and the High Minus Low (HML) portfolio, representing the
difference between returns on portfolios of value and growth firms. The relatively new
five-factor model (Fama and French, 2015) complements the three-factor model with
the Conservative Minus Aggressive (CMA) and Robust Minus Weak (RMW) spread
portfolios. CMA represents the return difference of a portfolio investing in firms with
conservative investment strategies minus a portfolio investing in firms with aggressive
investment strategies. RMW represents returns on firms with robust operating
profitability minus returns on firms with weak operating profitability.
The beta with respect to the index is 0.96; not surprisingly, close to 1. Importantly, the
relative momentum portfolio generates an alpha of 0.47 basis points (or 5.64% per
year), with a standard error of less than 15 basis points. The alpha is thus highly
statistically significant. Relative to the Fama-French three-factor model, the ESG
portfolio still generates an alpha of 0.47% per month, and the alpha remains statistically
significant. Adding two additional factors does not change this conclusion.
The SMB and HML loadings are not statistically significantly different from zero,
suggesting the ESG portfolio has neither a value nor a size bias. In the five-factor
model, the CMA exposure is borderline statistically significant and negative. The
While the alphas for the relative momentum portfolio are significantly different from
zero, the alphas for the absolute ESG momentum portfolios, reported in Panel B of
Table 1, are positive but no longer statistically significant. The factor exposures of the
absolute ESG momentum portfolio are very similar to those of the relative ESG
momentum portfolio.
Note: The analysis uses monthly returns. Standard errors are reported in
parentheses. The market model uses the MSCI index as the benchmark.
250
200
150
100
50
0
1-Jan-13 1-Jan-14 1-Jan-15 1-Jan-16 1-Jan-17 1-Jan-18
One possible explanation for this result is that the relative measure has more chance of
selecting firms that have low absolute ESG scores, i.e., firms that may be less likely to
be on investors’ radar screens as potential ESG target firms. However, it also raises the
possibility that the selected firms may not rank very high on ESG performance in an
absolute sense. Indirect evidence addressing this issue is presented in the next section.
10
While corporations now largely self-report some sustainability data, due to the lack of
standards and metrics, significant 'green-washing' and self-reporting data biases, ESG
scores contain a significant amount of noise and thus are of limited use for investment
purposes. In fact, typically, companies carry out voluntary reporting on their
sustainability performance in order to assure their shareholders and investors of their
compliance to regulations (Braam and Peeters, 2017). However, as more companies
are wary of the adverse impact of negative sustainability performance on investor
decisions, they may fail to disclose negative information (Reimsbach and Hahn, 2013).
A useful complement to the reported sustainability data, is Big Data leveraging
Artificial Intelligence technologies to extract, process, and analyze large-scale
structured and unstructured data on ESG and SDG-related factors, which can then
enable the integration of these sustainability factors into the decision-making of global
11
There are scenarios in which the technology can go wrong or provide imperfect
information; relying on publicly available information such as newspaper articles, may
lead to false or biased scores, for example. Other issues include fake news, articles that
commemorate negative events from the past, major discrepancies between reported
and third-party data, etc. For these reasons, it is necessary to perform extensive manual
verification of data to evaluate if the analysis corresponds to reality and implement
preventive measures. Extreme scores should be further examined using the underlying
data sources.
In this paper, we use Global AI Corp.’s (GAI) specific SDG scores. The company
extracts, filters, and cleans massive amounts of structured and unstructured data,
including self-reported company data, news articles, blogs, NGO reports, social media,
etc. to provide “raw,” short-term and long-term scores. The full data set covers
information across 60 languages from more than 100 countries. Specialized algorithms
map the raw data to specific companies and associated entities, such as subsidiaries,
using different combinations of company names, abbreviations, tickers, and ISINs.
Proprietary technology then ranks and filters content by relevance using domain-
specific taxonomies based on the SDGs.
The algorithms analyze the filtered content at a daily level: recording the number of
relevant news items, providing a sentiment score per news item, and tracking volume
and dispersion of sentiment across news items. This information is aggregated into
daily, company-specific “raw” scores, which represent aggregate sentiment of the SDG
data. GAI then aggregates data from 7 days of information, using statistics on the
precision of the scores and the volume of the news sources, accommodating sparsity in
the data while weighting recent information more heavily. For each company scores are
available for all 17 SDGs, and the system also provides an overall company score
measuring the overall SDG footprint of a company. The scores can be interpreted
roughly as “z-scores,” varying mostly between -1 and +1, and having a standard
12
The higher the score, the more positive the news is in relationship to each SDG, and
vice versa. For example, for SDG 13 (climate action), a company would get a more
negative score after a chemical spill that pollutes the ecosystem than a company that
increases its carbon emissions by 5%. The combination of positive and negative SDG
scores can be used to better assess non-financial risks and calculate a 'net' SDG
footprint that measures the effect of positive and negative externalities at both long-
and short-term frequencies.
We use GAI’s data across the MSCI US index universe over the Jan-2015-Dec. 2019
period to measure the SDG footprint of the active portfolio relative to the benchmark.
For this purpose, we apply the portfolio and benchmark weights to the SDG scores,
averaged for each year.
Our analysis addresses two different questions. Firstly, we verify whether ESG
momentum relative to the benchmark coincides with positive SDG footprint in the year
the ESG momentum was detected for the active portfolio constituents. In other words,
we test whether an ESG momentum strategy selects firms with an SDG footprint that
is better than that of the benchmark. Secondly, we investigate the SDG footprint of the
selected companies in the investment year (the year after ESG momentum was
observed). This exercise measures whether firms with ESG momentum continue to
relatively improve their SDG footprint in the year after their ESG scores increased and
whether ESG momentum is associated with a persistent (relative) positive SDG
footprint. Neither question needs to necessarily receive a positive answer. Because the
SDG scores are relatively fast moving, it is conceivable that they pick up certain ESG
issues even before the MSCI ESG rating change occurs. Unless companies continue to
generate (relatively) positive SDG contributions for a few years, it may not show up in
our measurement.
13
14
The last plot in Figure 2 averages the scores over the three years. Averaged over all
three years, the SDG footprint of the portfolio is better than the footprint of the
benchmark for all SDGs except for Gender Equality, SDG 5. The differences are
relatively small, however, and most of the scores (14 out of 17) are negative. This is not
15
16
Figure 4:
17
Statistical Significance
The polar plots show that the SDG footprint of the ESG portfolio is better both in the
year ESG momentum was observed (“contemporaneous”) and the subsequent
investment year. We now verify whether the differences are statistically significant. The
lack of observations prompts us to increase statistical power by comparing SDG
footprints on a monthly basis. For the contemporaneous comparison we have 3 years
of data, or 36 monthly observations; for the investment year we have 4 years, or 48
monthly observations. A simple t-test is performed to address whether the average
difference between the monthly SDG footprint of the portfolio and the benchmark is
statistically significantly different from zero. These observations may be serially
correlated which we control for by using 6 Newey-West (1987) lags in the creation of
our standard errors.
Table 2 reports the results, both for the raw scores and the short-term SDG scores. As
observed from the polar plots, all differences are positive, indicating that the SDG
footprint of the portfolio is better than that of the benchmark. Moreover, these
18
Table 2:
IV. Conclusion
Assessing company performance regarding ESG issues and SDG fitness profile is
challenging for the investors, academia, and NGOs. Because companies with good
ESG performance may enjoy a valuation premium, ESG investing has been thought
to create a potential conflict for asset owners who have a fiduciary duty not to
sacrifice long-term return opportunities. In this paper we dispel that view. We
investigate the SDG footprint of an active ESG portfolio using algorithms and
alternative data. We show it is feasible for an asset owner to both uphold his/her
fiduciary duty and have a positive impact on achieving the SDGs.
19
There is hope that Big Data can help investors better understand the underlying risks
in corporate behavior, and ultimately, make more sustainable investment decisions.
Objective data on the SDG footprint of companies and countries may also contribute
to better policy making regarding the realization of the SDGs. In future work, we
will also verify the relationship between financial returns and SDG footprint directly,
rather than through the narrower ESG lens.
20
Antoncic, Madelyn (2019). “Why sustainability? Because risk evolves and risk management
should too”, Journal of Risk Management in Financial Institutions, 12 (3): / Number 3 206-216.
Antoncic, Madelyn (2020). “Uncovering hidden signals for sustainable investing using
Big Data: Artificial intelligence, machine learning and natural language
processing,” Journal of Risk Management in Financial Institutions, 13 (2), 106-113
Bennani, Leila , Le Guenedal, Theo, Lepetit, Frederic, Ly, Lai , Mortier, Vincent,
Roncalli, Thierry, Sekine, Takaya (2018). “How ESG Investing Has Impacted the
Asset Pricing in the Equity Market,” AMUNDI Asset Management, available SSRN
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3316862
Berg, Florian, Julian Koelbel and Roberto Rigobon (2019). “Aggregate Confusion:
The Divergence of ESG Ratings,” Working Paper.
Braam, G. and R. Peeters (2017). “Corporate Sustainability Performance and
Assurance on Sustainability Reports: Diffusion of Accounting Practices in the Realm
of Sustainable Development,” Corporate Social Responsibility and Environmental
Management 25(2), 164-181.
Bril, Herman. (2020). “The Imperative of Sustainable Investing for Asset Owners.” In
H. Bril. G. Kell and A. Rasche (eds.). Sustainable Investing: A Path to a New
Horizon. London et al.: Routledge (forthcoming).
Duffie, Darrell (2010). “Asset Price Dynamics with Slow-Moving Capital” (American
Finance Association Presidential Address), Journal of Finance 65, 1238-1268.
Fama, Eugene F. and French, Kenneth R. (1993). “Common Risk Factors in the
Returns on Stocks and Bonds,” Journal of Financial Economics 33, 3-56.
21
MSCI (2019). “MSCI ESG Ratings Methodology: Executive Summary,” MSCI ESG
Research, available at
https://www.msci.com/documents/1296102/14524248/MSCI+ESG+Ratings+Met
hodology+-+Exec+Summary+2019.pdf/2dfcaeee-2c70-d10b-69c8-
3058b14109e3?t=1571404887226
Newey, Whitney K., and Kenneth D. West (1987). “A Simple, Positive Semi-
Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix,”
Econometrica, 55(3), 703-708.
RBC Global Asset Management (2019). “Responsible Investment Survey.” Available at
http://go.pardot.com/l/441592/2019-10-14/qbhs24
Reimsbach, D. and R. Hahn (2013). “The Effects of Negative Incidents in
Sustainability Reporting on Investors’ Judgments-an Experimental Study of Third-
party Versus Self-disclosure in the Realm of Sustainable Development,” Business
Strategy and the Environment 24(4), 217-235.
Serafeim, George (2020). “Public Sentiment and the Price of Corporate
Sustainability,” Financial Analysts Journal 76(2), 26-46.
United Nations (2006). “Secretary-General Launches ‘Principles for Responsible
Investment’ Backed by World’s Largest Investors,” Press Release, April 27. Available
at https://www.un.org/press/en/2006/sg2111.doc.htm
United Nations (2019), “Guidance on core indicators for entity reporting on
contribution towards implementation of the Sustainable Development Goals”,
22
23
24
4. Portfolio Rebalancing:
a. The selected stocks will remain in the portfolio for 1 year until the next
rebalancing date (i.e. 31 Dec of next year).
25
26
27