Content uploaded by Thierry Roncalli
Author content
All content in this area was uploaded by Thierry Roncalli on Nov 26, 2020
Content may be subject to copyright.
ESG Investing in Recent Years:
New Insights from Old Challenges∗
Angelo Drei, Th´eo Le Guenedal, Fr´ed´eric Lepetit,
Vincent Mortier, Thierry Roncalli & Takaya Sekine
Amundi Asset Management, Paris
November 2019
Abstract
This research is an update of the study that we published last year (Bennani
et al., 2018) and that explored the impact of ESG investing on asset pricing
in the stock market. It extends the original period 2010–2017 by adding eigh-
teen months from January 2018 to June 2019. These new results confirm the
previous results as we reach the same essential conclusions once again. ESG in-
vesting tended to penalize both passive and active ESG investors between 2010
and 2013. Contrastingly, ESG investing was a source of outperformance from
2014 to 2019 in Europe and North America. Moreover, ESG can be considered
as a risk factor in the Eurozone, while it continues to be an alpha strategy in
North America. However, the last 18 months exhibit new interesting patterns.
First, we observe a transatlantic divide since the results for North America
and the Eurozone are different for the recent period. Second, we document a
partial ordering between ESG ratings and performance that can be explained
by a shift from a static to a dynamic approach to ESG investing. Third, we
note some discrepancies between active and passive management. Fourth, the
Social pillar seems to have gained traction these last years, and is no longer the
laggard pillar. Fifth, factor investing and ESG investing are more and more
connected. In what follows, we develop and explain these five key findings.
Keywords: ESG, environmental, social, governance, asset pricing, active
management, passive management, factor investing.
JEL classification: G11, M14, Q01.
∗Angelo Drei, Th´eo Le Guenedal, Fr´ed´eric Lepetit, Thierry Roncalli & Takaya Sekine
work in the Quantitative Research department at Amundi Asset Management, and Vincent
Mortier is the Amundi Group Deputy CIO. We would like to thank Alice De Bazin, Mohamed
Ben Slimane, Tegwen Le Berthe, Philippe Ithurbide, Elodie Laugel and Caroline Le Meaux
for their helpful comments and discussion.
1
ESG Investing in Recent Years: New Insights from Old Challenges
1 Introduction
In November 2018, we published a discussion paper1“How ESG Investing Has
Impacted the Asset Pricing in the Equity Market”, which summarized the com-
prehensive research “The Alpha and Beta of ESG Investing” of Bennani et al.
(2018a). Leveraging our in-house ESG dataset and focusing on three method-
ologies — active management, passive management and factor investing —
we concluded that there was a radical break around 2013–2014, with greater
integration of ESG criteria in North America and Eurozone stock markets be-
tween 2014 and 2017, but the impact of ESG screening on return, volatility
and drawdown was still highly dependent on the selected time period and the
investment universe. More specifically, we took note of two success stories be-
tween 2014 and 2017, namely the Environmental pillar in North America and
the Governance pillar in the Eurozone, and one Japanese puzzle, where the
2010–2013 period was more favorable for ESG screening than the 2014–2017
period. Finally, our factor investing framework showed that ESG could be con-
sidered as a risk factor in the Eurozone, showing advanced ESG integration in
that area, but it still remains an alpha strategy in North America.
One of the motivations behind this research was to offset a contention with
academic findings. Many studies are based on long-term historical data, while
the tools created for the extra-financial analysis of companies are recent and
the sustainability investment space is quickly evolving. ESG regulations and
sustainable investing funds have both increased drastically, creating specific
market conditions that cannot be tested too far back in time. In consistency
with this reasoning, this discussion paper does not shy away from monitoring
trends in the ESG space over a short period of time, and focuses on the last
eighteen months2from January 2018 to June 2019, while our 2018 discussion
paper focused on the 2010–2017 period. Despite this short time frame, we
still identify some interesting trends that we believe should help sustainable
investors to better implement ESG in their portfolios.
First, we notice a major divergence between America and Europe in ESG
equity trends. While these two regions (accounting for more than three quar-
ters of the MSCI World’s weight) both showed positive advancements in ESG
integration in our previously studied period, we now observe a setback in North
America but some progress in the Eurozone, particularly on the active man-
agement side. The returns of North American long-short portfolios are less
than in the previous 2014–2017 period for all dimensions, and even slightly
negative on the Environmental pillar. Being an environmental investor during
1The discussion paper and the associated working paper are available in the Amundi
Research Center: http://research-center.amundi.com/page/Article/2019/01/The-
Alpha-and-Beta-of- ESG-investing.
2The new period is denoted by 2018–2019 even if it only concerns the first half of 2019.
2
ESG Investing in Recent Years: New Insights from Old Challenges
this last period was a losing bet. On the other side, the Eurozone gains even
more momentum on some pillars and stays positive for all long-short portfolios,
hence the idea of a halt in the convergence of these two investment universes,
or a transatlantic divide.
We also observe a development in approaching ESG integration. We in-
terpret the positive returns of our long-short portfolios (built as being long
on the top quintile and short on the bottom quintile) as a proof that best-
in-class selections and worst-in-class exclusions are still acceptable approaches
to ESG investing. In the new period 2018–2019, the Q1−Q5performance
is still positive on all dimensions in North America and the Eurozone, with
the exception of the Environmental pillar in North America. However, look-
ing at intermediary portfolios, especially the Q4quintile stock selection, we
cannot ignore that there is more to the Q1−Q5story. We hypothesize the
seemingly abnormal performance of Q4as the emergence of forward-looking
strategies, with some investors betting on improving companies rather than
well-scored companies. This shift towards a dynamic view is still a positive
development, as this increase in complexity of ESG integration demonstrates
that sustainable investors might better understand underlying issues and are
moving beyond a binary black and white view of corporations.
On the passive management side, we note a reduction in the excess return.
This development is the consequence of both the better integration of ESG
pillars in market pricing and the shift towards dynamic views as illustrated
through the Q4puzzle question. However, the tracking error cost of increas-
ing portfolio scores is stable overall compared to the previous period and we
still measure a noticeable tracking error cost premium for optimizing on the
Governance pillar compared to the Environmental and Social pillars.
In the 2018 paper, we observed a delayed integration of the Social pillar.
We acknowledged that while the Environmental and Governance pillars had a
turning point around 2013/2014, the integration of the Social pillar came later
around 2016. Therefore, it is without too much surprise that we observe this
pillar performing very well in both sorted and optimized portfolios during the
2018–2019 period. We draw a parallel between trends in equity markets and so-
cial narratives both within and outside the corporate landscape. For instance,
we think about the U.S. Business Roundtable public statement declaring it
“redefines the purpose of a corporation to promote an economy that serves all
Americans” or the awarding of economists whose research focused on alleviat-
ing global poverty.
The results of our factor analysis remain practically unchanged since the
last publication. We still view ESG as an alpha strategy in North America,
while in the Eurozone, it is the best explaining single-factor of stock returns
and makes a lot of sense to be included in a factor investing portfolio.
3
ESG Investing in Recent Years: New Insights from Old Challenges
2 The Transatlantic Divide
As a proxy for active management strategies, we have implemented the Fama-
French method of sorted portfolios. On a quarterly basis, we rank the stocks
based on their score, and build five sector-neutral quintile portfolios3. Portfo-
lio Q1corresponds to the 20% best-ranked stocks, whereas portfolio Q5corre-
sponds to the 20% worst-rated stocks. The selected stocks are then equally-
weighted and each portfolio is invested on the last trading day of the previous
quarter, at the closing price, and is held for three months4.
Looking at the Q1−Q5long-short portfolios in North America (Figure
1) and the Eurozone (Figure 2), we can see the evolution of the integration
of ESG and its subdimensions in both markets5. In the 2010–2013 period,
sustainable investors were penalized, as seen by the negative return of the
Q1−Q5long-short portfolios. In 2014–2017, after the radical break in ESG
integration, which we discuss in greater detail in our 2018 discussion paper,
ESG investing gained momentum and yielded positive returns on all pillars on
both sides of the Atlantic.
After eight years of parallel development, we observe a contradictory trend
in ESG investing between North America and the Eurozone since 2018. In-
deed, the last period 2018–2019 is marked by a squeeze in long-short returns
on all dimensions in North America, and even a loss on the Environmental
pillar6. This loss is important because it is the first long-short portfolio with
a negative return since the 2013/2014 ESG turning point in these two invest-
ment universes, and it reminds us how much the performance of ESG investing
is also regulation-driven. In this case, our first intuition is to look at the shift
in American public policies, notably the announced withdrawal of the United
States from the Paris Climate Agreement and some of the changes at the U.S.
Environmental Protection Agency (EPA), which might have impacted the as-
set pricing of the Environmental pillar. However, the negative performance of
the Environmental pillar should not overshadow the performance reduction of
the Social and Governance pillars during the 2018–2019 period, meaning that
the concerns relate to the three pillars of ESG investing.
On the Eurozone side, the verdict is more positive. All long-short portfolio
returns are positive. During the 2018–2019 period, the Environmental and
Social pillars yield even stronger returns comparatively to the previous pe-
riod. The decline of the Governance long-short portfolio return can be partly
3Given a universe of stocks, each portfolio is composed of 20% of the assets.
4The full methodology is explained in Bennani et al. (2018a).
5The portfolios are built using the MSCI North America and MSCI EMU indices as the
respective equity universes.
6The annualized return is equal to −0.8% for the 2018–2019 period whereas it was equal
to +4.1% for the previous period 2014–2017.
4
ESG Investing in Recent Years: New Insights from Old Challenges
Figure 1: Annualized return of Q1−Q5long-short portfolios (North America)
ESG Environmental Social Governance
−2
0
2
4
Return (in %)
2010 - 2013
2014 - 2017
2018 - 2019
Figure 2: Annualized return of Q1−Q5long-short portfolios (Eurozone)
ESG Environmental Social Governance
−2
0
2
4
6
8
Return (in %)
2010 - 2013
2014 - 2017
2018 - 2019
5
ESG Investing in Recent Years: New Insights from Old Challenges
attributed to a mean-reversion effect after an extraordinary period of impres-
sive performance. Indeed, the annualized return was 7.9% between 2014 and
2017, compared to just 1.3% for the most recent period. It follows that, if we
consider the screening based on the global ESG score, the recent period is in
line with the previous period since the two periods post an annualized return
around 6%.
Figure 3: Number of ESG regulations
1990 1995 2000 2005 2010 2015 2020
Year
0
50
100
150
Number of regulations
176
59
29
21
Europe
Asia
Latin America
North America
Source: PRI, responsible investment regulation database, 2019.
How to explain the transatlantic divide? In Bennani et al. (2018b), we
assumed that two main effects contributed to the ESG performance from 2014
to 2017: the selection effect of ESG screening and the demand effect of ESG
screening. By selection effect, we think about the direct impact of extra-
financial information on stock prices. By considering other risk dimensions,
the ESG investor may select corporations that are better managed from social,
environmental and governance points of view, or may avoid corporations that
present extra-financial weaknesses. The underlying idea is that sooner or later
these extra-financial risks have a financial impact on the performance of the
corporation. The second effect is related to the supply/demand balance. In-
deed, a stock price is the equilibrium between the supply and the demand for
this stock. It is obvious that ESG investment flows that have been observed
these recent years may have largely contributed to the good performance of
6
ESG Investing in Recent Years: New Insights from Old Challenges
ESG investing over the 2014–2017 period. For the most recent period, we may
think that European investors’ serious interest in ESG issues continues to im-
pact supply and demand with a subsequent effect on European stock prices in
2018 and 2019. For North American equities, the mobilization of European
investors has been followed by relatively strong engagement by Canadian in-
vestors, but a relatively weak implication of U.S. investors. A first explanation
of the American setback can be found in these engagement differences. But as
explained before, another justification could be the public policy of the Trump
administration in terms of its ESG roadmap. It follows that a third component
may contribute to the ESG performance: the political and regulatory environ-
ment. In Figure 3, we report a summary of the regulatory perspective between
Europe, North America, Asia and Latin America. For the last few years, we
observe that ESG regulations have gained the greatest momentum in Europe,
while North America has recently stalled, and we believe these developments
are linked to the transatlantic divide7.
3 A New Paradigm and the Q4Puzzle
When plotting the annualized return of the sorted portfolios, we notice an
anomaly on some Q4portfolios. Notably, for Social sorted portfolios in North
America (Figure 4) and Environmental sorted portfolios in the Eurozone (Fig-
ure 5), we observe a performance of Q4that is at Q1’s level or above. Curiously,
this ranking disorder only happens for the Q4sorted portfolio.
We refer to this phenomenon as a puzzle because it goes beyond the bi-
nary outcome in which Q1Q5holds or does not. In these two specific cases
(Social in North America and Environmental in the Eurozone), we still ob-
serve a positive return for the Q1−Q5long-short portfolio, so the abnormality
of Q4does not point towards the end of best-in-class and worst-in-class ap-
proaches. These strategies are still viable and yield respectively 1.6% and 5.4%
in annualized returns during the last period. Rather, we believe that the Q4
puzzle marks the emergence of new ESG investment strategies. The Q1−Q5
approach is representative of a static view of ESG scores, when best-in-class
stocks remain best-in-class stocks and worst-in-class stocks remain worst-in-
class stocks, while playing intermediary quintiles, especially the fourth quintile,
seems to be related to the strategy of ESG improvers or ESG momentum and
a dynamic view of ESG scores. During the last two years, ESG strategies
have become more complex, and this may explain the ranking disorder of the
Q4portfolio.
7In particular, through the last years, we observe that the slope of the European curve,
that includes the regulations at the country level, is stronger than the North American slope,
that includes the regulations at the state level for the United States.
7
ESG Investing in Recent Years: New Insights from Old Challenges
Figure 4: Annualized return of sorted portfolios (North America, Social, 2018–
2019)
Q1Q2Q3Q4Q5
Sorted portfolio
4
5
6
7
Return (in %)
5.7
4.6
4.3
7.3
4.1
Q1to Q5
Avg. of Q1,2,3
Figure 5: Annualized return of sorted portfolios (Eurozone, Environmental,
2018–2019)
Q1Q2Q3Q4Q5
Sorted portfolio
-2
0
2
4
Return (in %)
3.2
-2.2 -2.3
3.1
-2.1
Q1to Q5
Avg. of Q1,2,3
8
ESG Investing in Recent Years: New Insights from Old Challenges
We also note greater stabilization of ESG scores at the extremities over
time in our dataset. This means that, in the last few years, well ranked compa-
nies are even more likely than before to continue to be well viewed by the ESG
analysts, and similarly for poorly ranked companies. This stickiness might
have forced some investors to look for other opportunities in more dynamic ar-
eas, namely middle quintile stocks. Additionally, our performance attribution
analysis showed that the return of well-performing Q4portfolios is mostly due
to selection effects.
This finding is in line with the results reported by the Global Sustainable
Investment Alliance (2019). In its 2018 investment review, the organization
documents that the most common way to participate in sustainable invest-
ing (as measured by assets under management allocated to each strategy) is
to implement negative screening, but this approach is closely tailed by ESG
integration8and corporate engagement strategies. Similarly, the European
Sustainable Investment Forum (2018) found similar results a year before, and
stated that “the main strategy is exclusion, but in the last two years the
growth rate of this strategy slowed down. In contrast, best-in-class and ESG
integration have had a high growth rate”. Investment strategies based on the
dynamics of ESG ratings do not clearly correspond to negative or positive
screening, but they are more related to ESG integration. In this approach, an
improvement of an extra-financial criterion may lead to portfolio rebalancing,
exactly as we observe for financial ratios. It is obvious that the convergence be-
tween the extra-financial approach and the traditional security analysis based
on financial statements would certainly help increase the focus on the dynamics
and momentum of ESG ratings and not just their static level.
4 Optimized Portfolios: Not As Easy As Be-
fore
Concurrently with the development of active strategies, ESG integration is also
commonly deployed in passive management using the tilted portfolio technique.
The popular approach is to select a benchmark representative of the universe
to which the investor wishes to gain exposure, then increase its ESG exposure
while controlling for the tracking error volatility of the managed portfolios.
More frequently, investors will define additional constraints to limit the port-
folio turnover or sector-related bets. For this research, we simply try to find
the minimum tracking error portfolio for a given increment in ESG score.
8This is defined by the Global Sustainable Investment Alliance as “the systematic and
explicit inclusion by investment managers of environmental, social and governance factors
into financial analysis”.
9
ESG Investing in Recent Years: New Insights from Old Challenges
For each level of score improvement, this approach gives us two metrics: the
corresponding tracking error and the annualized excess return.
In Figure 6, we report the relationship between the ESG excess score and
the tracking error for the global DM universe9. For instance, targeting an
improvement of 0.5 for the ESG score requires accepting a tracking error of
40 bps. We obtain similar results to the ones we found in the 2018 discussion
paper in terms of tracking error level. Investors must accept a tracking error
risk if they want to implement ESG in a passive management framework, where
the benchmarks correspond to market capitalization-weighted (CW) indices10.
This clearly raises the definition of a strategic asset allocation (SAA) policy.
The fundamental issue is the reconciliation between ESG investing and SAA
based on CW indices when institutional investors are sensitive to the tracking
error risk. And, most of the time, they are!
Figure 6: Efficient frontier of optimized portfolios (Global DM)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Excess score
25
50
75
100
125
Tracking error (in bps)
ESG
E
S
G
Let us now consider the second important dimension of ESG tilted portfo-
lios. In Figure 7, we report the relationship between the excess score and the
excess return. For instance, an excess score of 0.5 led to an annualized excess
return of +22 bps for the 2018–2019 period. During the 2014–2017 period,
we found that the same excess score generated an excess return of +16 bps.
9It corresponds to the investment universe (also called index universe or equity universe)
of the MSCI World index.
10Moreover, we confirm that the Governance pillar generates more tracking error than
the two other pillars.
10
ESG Investing in Recent Years: New Insights from Old Challenges
Results are therefore similar. As previously, we also notice that the relation-
ship between excess score and excess return is not necessarily monotonous. It
increases when we target a low excess score and decreases when we target a
high excess score. This reversal phenomenon is most likely due to the diversi-
fication effect. Indeed, by increasing the excess score, we reduce the number
of positions held in the managed portfolio. Therefore, there comes a threshold
where the gains from the ESG screening are offset by the losses resulting from
the diversification reduction.
Figure 7: Annualized excess return of optimized portfolios (Global DM)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Excess score
−60
−40
−20
0
20
Excess return (in bps)
ESG
E
S
G
For a given universe and a specific score (E,S,Gor ESG), we can compute
the maximum excess return (MaxER) for the period11 and the reversal excess
score (RevES) beyond which the excess return becomes negative. For instance,
in the case of the global DM universe and the ESG score, we can see that up
to 0.8, the excess return remains positive and the maximum excess return is
equal to +22 bps and is reached when the excess score is equal to 0.5. It follows
that MaxER = 22 bps and RevES = 0.8.
Comparing our optimized portfolios on North American (Figure 8) and
Eurozone investment universes (Figure 9) might give the wrong impression
that it invalidates our transatlantic divide thesis. North American portfolios
(except for the Environmental pillar) follow the representative concave shape
we expect when running this kind of optimization. On the contrary, Eurozone
11It corresponds to the highest excess return among all optimized portfolios.
11
ESG Investing in Recent Years: New Insights from Old Challenges
Figure 8: Annualized excess return of optimized portfolios (North America)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Excess score
−150
−100
−50
0
50
Excess return (in bps)
ESG
E
S
G
Figure 9: Annualized excess return of optimized portfolios (Eurozone)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Excess score
−50
0
50
100
150
Excess return (in bps)
ESG
E
S
G
12
ESG Investing in Recent Years: New Insights from Old Challenges
portfolios break this pattern. In fact, the results for the Eurozone universe
seem at first sight inconsistent with respect to those obtained in the case of
sorted portfolios.
To better understand these seemingly conflicting results, we dive into the
relationship between size and ESG, and notice an important effect. For large
and mega-cap12 stocks, ESG sorting generates a performing selection effect
across the whole scoring range, or, Q1BLQ5where BLis the equally-
weighted portfolio of large-cap stocks. For mid-cap stocks, ESG filtering seems
to work well on the exclusion side but not from a best-in-class perspective, and
we have the less favorable following setting BMQ5Q1, where BMis the
equally-weighted portfolio of mid-cap stocks. Unfortunately, our unconstrained
optimizations generate portfolios that continuously overweight these mid-cap
Q1stocks. We believe this is related to scoring issues: noisiness in the ESG
signal due to lower volumes of information, potentially inappropriate decay
parameters, focus on average ratings as opposed to score confidence intervals.
This extends our new paradigm discussion, with sustainable investors who will
have to develop more complex views of ESG, depending on company sizes.
Another reason is the previously mentioned loss of diversification. The
North American universe includes on average more than 700 stocks, while the
Eurozone universe is composed of about 240 stocks. This is directly reflected
in our optimization results. When we target an ESG excess score of +1.0,
the North American portfolio holds on average 175 stocks, while the Eurozone
portfolio contains about 100 stocks. Another aspect to take into account is the
impact of the Q4puzzle on optimized portfolios. This creates more instability
even when the return of the Q1−Q5long-short portfolio is positive.
Table 1: Summary of optimized portfolios
Global DM North America Eurozone Japan
ESG E S G ESG E S G ESG E S G ESG E S G
2014–17
Maximum
ER 38 129 7 87 18 129 1 28 282 250 86 440 49 0 6 191
Reversal
ES 1+ 1+ 0.8 1+ 1+ 1+ 0 .7 1+ 1+ 1+ 1+ 1+ 0 0 1+
2018–19
Maximum
ER 22 8 33 25 42 5 41 36 0 29 184 0 0 24 0 0
Reversal
ES 0.8 0.5 0.8 0.9 0.7 0 0.6 0.7 0 0.8 1+ 0 0 0.7 0 0
We summarize our results on optimized portfolios13 in Table 1. We note a
reduction in both the maximum excess return and the reversal excess score for
12We classify large and mega-cap stocks as stocks with market capitalizations above e15
bn.
13With the addition of the Japanese investment universe, which corresponds to the stocks
of the MSCI Japan index.
13
ESG Investing in Recent Years: New Insights from Old Challenges
most pillar-universe pairs. In the Eurozone, where the loss of diversification
is reached faster than in North America, optimized portfolios generate poorer
results (except for the Social pillar). Overall, we think that tilted portfolios
can still capture excess returns but return profiles are less interesting than
before. This is in great part due to the Q4puzzle and the mid-cap story14 and
passive investors are required to integrate dynamic views of ESG in order to
avoid this squeeze.
5 Social: From Laggard to Leader
In Bennani et al. (2018a), we already noted that despite the identification of
an ESG turning point in 2013/2014, pillars have been integrated into stock
prices at different paces. Namely, Environmental and Governance benefited
from a head start in that space. One hypothesis is to consider these two
pillars as more straightforward from a risk management perspective. Even
investors who do not integrate the ESG lenses into their investment processes
understand that there is a general trend towards more taxation in order to
curb negative externalities linked to corporate pollution, and good corporate
governance could be a hedge against management fraud and principal-agent
problems.
The social case has been less obvious to investors but recent changes in its
narrative might have helped this last pillar to finally gain momentum. Since
the financial crisis of 2007–2008, inequality has been brought back to the table.
To illustrate this trend, we can mention the book Capital in the Twenty-
First Century written by Piketty (2014), of which The Economist attributed
its exceptional success to “being about the right subject at the right time”.
More recently, the Royal Swedish Academy of Sciences has just awarded the
Nobel Memorial Prize in Economic Sciences to three economists “for their
experimental approach to alleviating global poverty”. This social momentum
goes beyond the academic realm. Even in the backdrop of the China–United
States trade war, the current U.S. administration has been vocal about wanting
American companies to focus on domestic job creation, and there has been an
effort on the corporate side to at least show that the message has been heard15
(Business Roundtable, 2019).
14If we impose the mid-cap neutrality on ESG optimized portfolios, only Gtilted port-
folios underperformed in the 2018–2019 period in the Eurozone, while E,Sand ESG opti-
mized portfolios presented a positive excess return. Therefore, it is obvious that the scoring
of mid-cap (and small-cap) corporations is an important topic of ESG investing.
15We think about some of the statements of the Business Roundtable, which brought
together the CEOs of some largest American corporations, that define the purpose of a
corporation and would make Freeman (2010) blush.
14
ESG Investing in Recent Years: New Insights from Old Challenges
Figure 10: Annualized return of sorted portfolios (Eurozone, Social, 2010–
2017)
Sorted portfolio
Return (in %)
7.2
7.7
8.5
5.8
9.4
11.2
10.3 10.3
10.6
8.3
15.0
11.1
10.3
10.5
5.0
2010 -- 2013
2014 -- 2017
2016 -- 2017
Source: Bennani et al. (2018a).
Figure 11: Annualized return of sorted portfolios (Eurozone, Social, 2018–
2019)
Q1Q2Q3Q4Q5
Sorted portfolio
-6
-4
-2
0
2
4
6
Return (in %)
1.6
4.6
1.6
-6.7
-1.3
15
ESG Investing in Recent Years: New Insights from Old Challenges
Figure 10 is directly taken from our previous working paper (Bennani et al.,
2018a) and already shows well the gain in momentum of the Social pillar since
2016 in the Eurozone through annualized returns of sorted portfolios. Figure 11
is an update on the 2018–2019 period and shows that the Social pillar continues
to thrive. The Q1−Q5long-short portfolio posted a return of 2.9% whereas
a (Q1+Q2)−(Q4+Q5) long-short portfolio would display a performance of
7.1% if we consider the same long and short notional exposures! The fact
that Q1and Q2dramatically outperform Q4and Q5explains why Social is
the winning pillar when implementing passive management in the Eurozone
between 2018 and 2019. In North America, we have already presented the
returns of the sorted portfolios in Figure 4 on page 8. We had observed a
monotone increasing function between the Social score and the performance
except for the Q4sorted portfolio16 . However, this puzzle of the Social pillar
in North America is balanced by the good behavior of the pillar in the case of
optimized portfolios. In Figure 8 on page 12, we note that the Social pillar
exhibits good performance for low and medium excess scores. If we consider
the global DM universe, Social is clearly the winning pillar17.
Remark 1 Despite these good results, we have to be careful and not extrap-
olate these trends in the future. Indeed, we generally notice that the perfor-
mance of extra-financial criteria is mean-reverting and we observe some cycles
in ESG investing. For instance, Environmental was the winning pillar in Eu-
rope between 2010 and 2013. Then, Governance took the lead, while Social is
today on all fronts. In America, the cycle is less marked, but has begun with
Governance followed by Environmental. Since we also expect that more dy-
namic ESG investment strategies will be developed, it is extremely difficult to
be forward-looking. But it is obvious that in the past, the performance sequence
that we have observed can be related to the implementation sequence of ESG
investors.
6 All Quiet on the Factor Front
In the previous discussion paper, after finding evidence for greater ESG in-
tegration in the 2014–2017 period, we dared to wonder if ESG was a new
risk factor. To answer this question, we defined two main criteria for eligibil-
ity as risk factor: generating extra performance or reducing risk, and being
a complement to traditional risk factors. We concluded for the previous pe-
riod that ESG strategies remained alpha strategies in North America. They
16The annualized return is respectively 4.1% for Q5, 4.3% for Q3, 4.6% for Q2and 5.7%
for Q1.
17See Figure 7 on page 11.
16
ESG Investing in Recent Years: New Insights from Old Challenges
have generated outperformance, they are diversifying, but they cannot ex-
plain the dispersion of stock returns better than the standard five-factor risk
model based on size, value, momentum, low-volatility and quality. On the con-
trary, we drew a different conclusion for the Eurozone. We observed that ESG
improved diversification of multi-factor portfolios and concluded that in this
region, “an ESG strategy was more a beta strategy than an alpha strategy”
(Bennani et al., 2018b).
In order to test this hypothesis again, we use the same multi-factor analysis
and factor picking framework as those developed in Bennani et al. (2018a). For
the multi-factor analysis, we perform the cross-section regression of all stocks
that belong to North America and the Eurozone index universe with respect to
the CAPM, the five-factor model and a six-factor model that includes the ESG
risk factor18. Table 2 shows the average R2of each model. While the impact
of adding the ESG factor is positive for all periods and the two investment
universes, we concede that it has a minor impact in 2018–2019: 47.5% versus
46.3% in North America and 45.4% versus 44.2% in the Eurozone. However,
these results were roughly the same for the 2014-2017 period.
Table 2: Cross-section regression (average R2)
North America Eurozone
Period 2014–2017 2018–2019 2014–2017 2018–2019
CAPM 26.2% 34.9% 37.7% 32.6%
5F 35.4% 46.3% 45.3% 44.2%
6F (5F + ESG) 36.8% 47.5% 46.0% 45.4%
In addition to these linear regressions, we run lasso regressions with differ-
ent levels of penalty. The main advantage of this method is that it shows a
sequential ranking of factors for all levels of factor intensity19. This sequen-
tiality helps us to understand which factors contain overlapping information.
Figure 12 shows the results for North America. Quality is first selected, then
followed by ESG, momentum, low-volatility, and finally value20. Although the
18The five risk factors are the following standard factors: size, value, momentum, low-
volatility and quality. The factors are built similarly to our sorted portfolios and, contrary
to the academic literature, they correspond to Q1long-only portfolios, and not to Q1−Q5
long-short portfolios. The reason is that factor investing is massively implemented in long-
only portfolios. Therefore, it is more realistic to consider long-only risk factors in order to
conclude if ESG does make sense or not in a factor investing framework.
19A factor intensity of 0% corresponds to the lasso regression with a penalty so high that
no factor is selected, while a factor intensity of 100% corresponds to the lasso regression
with a penalty of 0, which is the same as the ordinary least square regression.
20We do not include the size factor since it is implicitly embedded in the other risk factors.
17
ESG Investing in Recent Years: New Insights from Old Challenges
Figure 12: Factor selection (North America)
Figure 13: Factor selection (Eurozone)
18
ESG Investing in Recent Years: New Insights from Old Challenges
ESG factor is the second factor selected, it ends up last once factor intensity
reaches 100%. This is interpreted as the ESG factor containing some informa-
tion in North America, but offering little diversification benefits in multi-factor
framework. In other words, ESG does make a lot of sense in a poorly diversi-
fied portfolio, but it does not improve the diversification in a portfolio that is
already well-diversified. This confirms our previous conclusion that ESG is an
alpha strategy for North America. Figure 13 concerns the Eurozone. In this
case, ESG is first selected, then followed by value, low-volatility, momentum
and finally quality. In this universe, the ESG factor holds better as the factor
intensity rises and ends up between low-volatility and quality when there is
no penalty. We confirmed our previous finding that ESG has become a beta
strategy in the Eurozone. In Bennani et al. (2018b), we already made an
analogy to the history of other traditional risk factors. One example was the
quality anomaly, which became popular after the Global Financial Crisis and
then stopped being an alpha strategy as it has been significantly invested in.
We believe something similar happened to the ESG factor in the Eurozone
during the 2014–2017 period and the last eighteen months of data confirm this
finding.
7 Conclusion
ESG investing is a rapidly evolving area. For instance, we report below some
figures from the latest report from the Global Sustainable Investment Alliance.
In 2016, the size of the global responsible investment market was $22.9 tn.
Two years later, it stands at $30.7 tn. With double-digit growth rates21,
ESG investing is certainly the most dynamic sector of the asset management
industry. This concerns all regions and all investment styles. It is obvious
that things are moving quite fast in such an environment. This was what
motivated us to update our previous research so quickly, because all players in
ESG investing have the perception that past results are no guarantee of future
results, especially in the socially responsible investing landscape.
Despite the short time period of our study, we come up with some important
findings — some positive, some disappointing — for sustainable investors.
Overall, it is a more complex environment we describe. These mixed results
are closely tied to the dual nature of ESG investing, as a financial innovation
(a risk model and a family of investment strategies) and as a moral view of
corporations.
ESG can be seen as an alternative risk model. ESG screening can in prin-
ciple provide a hedge against future pollution-related taxation, management
21The annual growth rate for the period 2014–2018 is 6% for Europe, 16% for the United
States, 21% for Canada, 50% for Australia/New Zealand, and 308% for Japan (GSIA, 2019).
19
ESG Investing in Recent Years: New Insights from Old Challenges
Figure 14: The market of ESG investing at the start of 2018
39%
6% 46%
USA
$ 12.0 tn
38% growth
in 2 years
JAPAN
$ 2.2 tn
(vs $ 474 bn in 2016)
CANADA
$ 1.7 tn
42% growth
in 2 years
AUSTRALIA / NZ
$ 0.7 tn
46% growth
in 2 years
EUROPE
$ 14.1 tn
11% growth
in 2 years
2%
7%
$ 30.7 tn
Global Responsible Investment
major markets in 2018
~ 2/5
of global assets under
management
+34%
growth in 2 years
Source: Global Sustainable Investment Alliance (2019).
fraud and other risks that cannot be derived from market data and accounting
figures alone. Our sorted portfolio results seem to point in that direction, but
the reduction in the ESG advantage in North America and the confirmation
that ESG is a beta strategy in the Europe tend to indicate that it is becoming
the norm rather than an advantage.
ESG investing is also the investment expression of what some investors
believe the world should be like. This separates ESG investing from other in-
vestment strategies in multiple ways. Financial gain is not always the expected
outcome of ESG investing22, and this probably explains why ESG investing has
emerged. As a consequence, ESG funds are bound to increase through the con-
version of investors who independently from financial expectations decide to
become sustainable investors. This creates a positive financial feedback loop
unique to the ESG factor, as demand-driven pressure should theoretically in-
crease the price of securities deemed appropriate for inclusion in a sustainable
investment strategy. This specific environment is beneficial to dynamic views,
and we see that there is already a reward for accompanying corporations on
their journey to the good side (as seen through our Q4puzzle and the rise of
22Berg et al. (2014) did not find “any significant added value, neither positive nor neg-
ative” in the 2005–2013 period, whereas Renneboog et al. (2008) found similar results for
previous periods.
20
ESG Investing in Recent Years: New Insights from Old Challenges
improvers, momentum and engagement strategies). However, this normativ-
ity is also a coordination game since moral views are not uniform. CSRHub
compared the ratings of two leading ESG rating agencies, MSCI and Sustain-
alytics, and found a correlation coefficient of 32% between the two datasets23 .
Some organizations, such as the network of investors behind the Principles for
Responsible Investment (PRI), try to increase the uniformity, but it is still a
work in progress. At a more extreme level, national governments can reject
these views and drastically affect the ESG risk pricing mechanism.
The fact that there are multiple views on ESG investing creates heterogene-
ity. In this situation, there are many ways to build a rating system. It is then
not unreasonable that ESG ratings are less correlated than credit ratings. In
fact, there are many issues about ESG ratings. First, they correspond to rel-
ative metrics, and not to absolute metrics. For instance, we can compare two
corporate entities from the same sector, but it is extremely difficult to compare
two firms that belong to different sectors or that have two different business
models. Second, the meaning of the ESG rating is unclear. In the case of
credit ratings, the question is: what is the one-year default probability of the
entity? An ESG rating does not answer a precise question, since it is a multi-
faceted metric that combines at least three dimensions: Environmental, Social
and Governance. The convergence can then only be obtained at more refined
levels. And it is illusory to think that we can obtain the same homogeneity at
the most aggregate level than we observe for the credit ratings. In fact, this
heterogeneity is inherent to the concept of ESG investing and questions the
interpretation of the ESG investing figures collected by GSIA (2019). Indeed,
according to Figure 14, North America competes on an equal footing with Eu-
rope. However, we know that there are strong differences between these two
regions in terms of ESG integration, whether it concerns asset owners or asset
managers.
In 2018, we were convinced that the break in 2014 marked a major shift
concerning the relationship between ESG investing and asset pricing in the
stock market and this radical change was irreversible. We believed that the
mobilization of investors on ESG issues was high enough to create a positive
bubble for ESG assets. This conviction was mainly explained by the increas-
ing interest of our clients. For instance, we have reported the frequency of
institutional requests for proposal received at Amundi in Figure 15. Of course,
these figures are biased because they only include an asset manager that is
highly active in ESG investing. However, even though these absolute figures
23No direct source. The fact was reported by Robin Wigglesworth in the Financial Times
on 20 July 2018. For comparison, this author mentions a correlation of 90% between credit
ratings assigned by S&P and Moody’s. However, we note that Bruder et al. (2019), using
the ESG database of MSCI, and despite some methodological differences, performed some
similar analyses and found results that do not differ much from ours.
21
ESG Investing in Recent Years: New Insights from Old Challenges
Figure 15: Frequency of institutional RFPs that require ESG filters
2010
2015
2017
2018
40%
20%
Source: Based on RFPs received at Amundi.
are not representative of the global asset management industry, we think that
the trend is very interesting to analyze. In 2010, the number of RFPs with
an ESG component was very low and less than 5%. In 2015, 20% of institu-
tional RFPs that were received at Amundi required an ESG filter. In 2017,
this figure was 40%. We clearly observe a trend and this trend is far from
over. Investor interest in ESG issues is certainly the major factor behind the
break in 2014. Indeed, we know that investment flows must be substantial in
order to impact supply and demand and then stock prices. However, this study
also shows that it is not sufficient. The regulatory environment and collective
political will are also another important component, because they give some
signals to the financial market. Without the right regulatory framework and
more transparency, we may face some ‘free rider’ or ‘moral hazard ’ problems24
This issue is also connected to the previous question about the heterogeneity
of ESG strategies and the definition of ESG investing.
However, while the ESG investing space is becoming more complex (envi-
ronmental policy reversal in the United States, shift from a static to a dynamic
view of ESG scores, lead/lag integration of the different dimensions), our re-
sults show that the ESG fundamentals are still present. Best-in-class and
worst-in-class approaches still work overall. This is good news on the invest-
ment side. On the financing side, an open issue remains: what is the impact of
ESG investing on the capital allocation and the cost of capital? This question
is left for forthcoming future research.
24Since ESG is a hot topic and also a marketing argument, we may observe some ‘green-
washing’ behavior or more simply the fact that being an ESG investor covers several aspects.
22
ESG Investing in Recent Years: New Insights from Old Challenges
References
[1] Bennani, L., Le Guenedal, T., Lepetit, F., Ly, L., Mortier, V.,
and Sekine, T. (2018a), The Alpha and Beta of ESG Investing, Amundi
Working Paper, 76, http://research-center.amundi.com.
[2] Bennani, L., Le Guenedal, T., Lepetit, F., Ly, L., Mortier, V.,
Roncalli, T. and Sekine, T. (2018b), How ESG Investing Has Impacted
the Asset Pricing in the Equity Market, Amundi Discussion Paper, 36,
http://research-center.amundi.com.
[3] Berg, F., De Laguiche, S., Le Berthe, T., Russo, A., and Sor-
ange, A. (2014), SRI and Performance: Impact of ESG Criteria in Eq-
uity and Bond Management Processes, Amundi Discussion Paper, DP-
03-2014, http://research-center.amundi.com.
[4] Bruder, B., Cheikh, Y., Deixonne, F., and Zheng, B. (2019), In-
tegration of ESG in Asset Allocation, SSRN,www.ssrn.com/abstract=
3473874.
[5] Business Roundtable (2019), Business Roundtable Redefines the Purpose
of a Corporation to Promote ‘An Economy That Serves All Americans’,
19 August 2019, www.businessroundtable.org.
[6] Eurosif (2018), European SRI Study 2018, 116 pages, November 2018,
www.eurosif.org.
[7] Freeman, R.E. (2010), Strategic Management: A Stakeholder Approach,
Cambridge University Press.
[8] Global Sustainable Investment Alliance (2019), 2018 Global Sustainable
Investment Review, GSIA Report, April, http://www.gsi-alliance.
org.
[9] Piketty, T. (2014), Capital In the Twenty-First Century, Harvard Uni-
versity Press.
[10] PRI (2019), Responsible Investment Regulation Map,www.unpri.org/
sustainable-markets/regulation-map.
[11] Renneboog, L., Ter Horst, J., and Zhang, C. (2008), Socially Re-
sponsible Investments: Institutional Aspects, Performance, and Investor
Behavior, Journal of Banking & Finance, 32(9), pp. 1723-1742.
[12] Wigglesworth, R. (2018), Rating Agencies using Green Criteria Suffer
from ‘Inherent Biases’, Financial Times, New York, 20 July 2018.
23
