Professional Documents
Culture Documents
Tblk-Elevation and Isometric
Tblk-Elevation and Isometric
PERFORMANCE
Erik Landry, Mariana Castillo-Lazaro, Anna Lee
INTRODUCTION
There are currently over 1500 Principles for Responsible Investment signatories with over $60 trillion in
assets under management. This growing number signals the rapidly increasing commitment to
responsible investing worldwide. Nevertheless, the investment community at large has traditionally
been cautious of having to trade-off between sustainability and alpha, or the performance of a managed
portfolio of securities relative to an appropriate benchmark. Consequently, Environmental, Social, and
Governance (ESG) metrics and scores – measuring corporate performance on issues like carbon output,
water usage, workforce turnover, or executive pay structure – are becoming ever more integrated into
investment decisions to make sure that they do not jeopardize return on investment. While the
relationship of ESG to alpha has been well studied in equity markets, the same cannot be said for credit
markets.
The Securities Industry and Financial Markets Association estimates that at the end of 2016, the debt
outstanding in the US Corporate Bond Market was over $8.5 trillion. The corporate bond market could
be a particularly good fit for sustainable investing because it promotes long-term reliability over short-
term gains and shareholder pressure. While ESG might not be a strong driver of upside performance, it
could potentially reveal insight to the responsible or irresponsible management of a company, thereby
reducing downside risk. Elucidating a clear relationship between ESG and improved corporate bond
performance could influence investors to favor sustainable companies, encourage companies to become
more sustainable, and ultimately dispose of the long held belief that sustainability efforts and company
performance are necessarily in opposition with one another.
In this report, we attempt to discover whether the sustainability of a company influences the
performance of its corporate bonds. As measurements of sustainability, both aggregated and separated
ESG scores from two industry rating and analysis companies – Sustainalytics and MSCI – were used. As a
measurement of corporate bond performance, the Credit Default Swap (CDS) is used. The CDS is
essentially insurance on the bond in the event that it defaults, and as such, is relatively sensitive
measure of the reliability of the credit security. We search for relationship in the relativity of ESG
metrics and CDS, in the predictive quality of ESG on CDS, and finally in the predictive quality of the
difference in ESG scores from different scoring agencies on CDS.
A 2015 study by Barclays developed corporate bond portfolios of similar character but with different
performance in terms of company ESG scores in order to study the effect of ESG ‘tilts’ on portfolio
performance. It showed that those portfolios with positive ESG tilt resulted in a “small but steady
1
performance advantage”, including higher credit ratings and lower spreads (Desclée, Dynkin, Hyman, &
Polbennikov, 2016). It also evaluated several other phenomena including that Governance scores had a
stronger relationship to performance than Social scores, that issuers with high Governance scores had a
lower probability of experiencing a credit rating downgrade, and that these results remained relatively
similar when using either MSCI or Sustainalytics ESG scores, despite their different methods of
calculation.
A 2016 study done by MIT and Breckinridge Capital Advisors examined the relationship between ESG
scores and the option adjusted spread (OAS) of corporate bonds. They found, similarly to Barclays, that
ESG scores are “positively correlated with small, stable spreads in corporate debt markets” (Clubb,
Takahashi, & Tiburzio, 2016). Similar correlations were found for return on assets and leverage ratios.
However, the authors acknowledged the fact that their analyses relied heavily on correlation, and
should not be conflated to represent causation.
This paper seeks to build upon of the findings of these two reports. Firstly, the 2016 MIT report
assessed the covariances and correlations between ESG, E, S, and G ratings and the option adjusted
spreads of all of their companies in aggregate. However, given that E, S, and G operations can have
significant differences in importance within different sectors of the economy, it could be reasonably be
argued that this level of analysis lacks appropriate granularity. Thus, this paper will address that gap by
utilizing a similar analysis, but further broken down by sector in order to give a more granular look into
the relationships between sustainability and corporate bond performance. Secondly, while correlations
might give a first indication of a possible relationship, many other financial parameters have much
greater influence on CDS than ESG scores. Ordinary Least Squares (OLS) regression is used to control for
several of these other factors in order to examine the predictive capability of ESG scores on CDS.
Thirdly, the ESG scores of Sustainalytics and MSCI were found to be very loosely correlated. This paper
evaluates whether or not the difference in ESG scores between Sustainalytics and MSCI bears any
relationship to CDS.
The breakdown of the data universe into economic sectors revealed differences in strength ESG
correlation with CDS for different sectors. In particular, the healthcare, industrials, and information
technologies sectors seems to show stronger correlations. On the other hand, real estate seems to
show the opposite directionality. While OLS regression revealed some sector influence on CDS, it also
revealed certain limitations of the data. Given the number of companies under examination and lack of
necessary financial control factors, the sector-specific E, S, and G interactions could not be statistically
proven. Given the weakness of correlation between the Sustainalytics and MSCI ESG scores, the
absolute difference between the two scaled scores was tested for strength as a predictor of CDS. The
regression of CDS of on the difference, as well as on the S&P credit ratings and economic sector
revealed, again revealed a lack statistically significant results. Ultimately, the confidence of the
regressions were limited by the assumptions of a OLS regression tool, as well as the narrow width of the
data.
DATA
2
The ESG data was provided by Breckinridge Capital Advisors, a Boston-based investment advisory firm
focused on fixed-income securities and a leader integrating ESG into their investment decisions. The
ESG data includes Sustainalytics ESG scores for the Standard & Poor’s (S&P) 500 – an index of 500
companies with common stock listed either on the NYSE or NASDAQ stock exchanges – as well as ESG
scores from MSCI on 3876 companies. Sustainalytics and MSCI are both leaders in ESG research and
ratings. They provide investors with data and analysis on ESG and other sustainability performance
indicators in order to provide greater transparency in financial markets. CDS data for the S&P500 in
January, 2017 was collected with DataStream from Thomson Reuters.
Treatment of the data included sub-setting for the required columns, deleting rows containing NAs and
blank cells, eliminating duplicate entries, sub-setting to select only the data pertaining to the relevant
date, and using credit ratings to filter for investment grade securities. The International Securities
Identification Number (ISIN) is a unique number for each issuer’s corporate bond and was used to merge
the datasets. The Global Industry Classification System (GICS) was used to assign companies of a given
industry into an appropriate sector. Once the Sustainalytics ESG, MSCI ESG, and Thomson Reuters CDS
data are merged, the intersecting data set of investment grade companies is narrowed to only 203
companies, hereafter referred to as the data ‘universe’.
3
The Sustainalytics ESG scores are scored out of 100. The MSCI scores are done so out of 10. A scaled
MSCI score was calculated as follows:
/01234156%7',8'(97,&%
!"#$%&'()*,, = !"#$, ∗
/012341:5;<
This not only puts the MSCI score on a scale of 100, but also scales MSCI scores up because
Sustainalytics tends to rate its companies in the 50-70 range whereas MSCI tends to rate its companies
in the 40-60 range (out of 100). The Scaled Difference is the Sustainalytics ESG score for a given
company minus the scaled MSCI ESG score of that same company. The Absolute ESG Difference is
simply the absolute value of this difference. Note that the Scaled MSCI ESG score mean does not exactly
match the Sustainalytics ESG Score mean in Table 1 because the averages used in the calculation were
taken over all available companies, not just those that are investment grade, in order to get a slightly
better idea of how Sustainalytics and MSCI tend to rate companies.
Once the Sustainalytics ESG, MSCI ESG, and Thomson Reuters CDS data are merged, the intersecting
data set of investment grade companies is narrowed to only 203 companies.
The weighting of MSCI data is straight-forward; for a given industry, a set of weights are applied to the E,
S, and G scores, resulting in the Weighted Average ESG Score. This is not so for the Sustainalytics Total
ESG Score. The lack of transparency in calculation of this Total ESG Score introduces a source of
uncertainty when clustering and comparing researcher-defined groups of companies because the ESG
Score may have been calculated differently for each of a set of undisclosed Sustainalytics-defined
groupings. Nevertheless, we assumed relative overlap in industry classification and used the GICS to
group companies into appropriate economic sectors.
PHASE I
While covariance and correlations between ESG, E, S, and G scores with the CDS give a first
approximation for the relationship between corporate sustainability and CDS, it obscures the operation
of that relationship within the ESG distribution. In order to give more insight into how CDS is distributed
over the distribution of ESG scores, the companies were divided into four sets of quintiles according to
their ESG, E, S, and G scores, respectively. The average CDS was then calculated within each quintile.
Additionally, in order to explore the difference in importance between E, S, and G factors within specific
sectors, these quintile tables were produced for each of the GICS sectors.
As shown in Table 3 below, the quintile analysis showed relatively clear directionality, with the sets of
higher performing ESG companies generally having lower CDS.
Quintile 1 2 3 4 5
ESG 95.6 65.7 78.7 76.7 64.6
E 95.9 79.7 75.7 61.3 68.7
S 91.1 69.9 74.6 71.0 74.8
G 76.0 84.5 80.2 71.9 68.9
4
TABLE 3. AVERAGE CDS PER SUSTAINALYTICS ESG QUINTILES FOR THE DATA UNIVERSE
NOTE: QUINTILE 5 REPRESENTS THE HIGHEST PERFORMING ESG, E, S, AND G SCORES, RESPECTIVELY
As shown in Tables 4a-d, the Healthcare, Industrials, and Information Technology sectors show even
stronger relation between ESG performance and average CDS. By contrast, the Real Estate Sector, at
least on the extremities, seemed to exhibit the opposite relationship.
Quintile 1 2 3 4 5
ESG 71.02 46.1325 45.6925 54.7475 23.2925
E 60.6 56.1275 38.03 42.4025 46.33
S 58.212 57.6875 63.365 41.53 23.2925
G 67.58 58.685 51.8975 30.4425 33.14
(a) Healthcare
Quintile 1 2 3 4 5
ESG 71.13143 62.53429 52.145 37.77286 50.28833
E 80.68714 55.95143 40.83333 46.91286 47.46833
S 56.25714 66.66714 60.435 41.63 50.03
G 69.18857 52.48571 57.165 56.83429 37.02
(b) Industrials
Quintile 1 2 3 4 5
ESG 106.9367 102.4633 82.35667 70.33333 67.555
E 106.9367 102.4633 64.75667 83.69333 73.915
S 106.9367 110.5367 74.28333 70.33333 67.555
G 102.6867 73.78667 103.26 82.35667 67.555
(c) Information Technology
Quintile 1 2 3 4 5
ESG 87.99333 100.2667 135.055 130.6933 127.045
E 87.99333 123.2167 83.81 141.9067 127.045
S 97.78667 160.4967 93.58 118.1233 82.34
G 87.99333 118.3233 121.625 142.8833 95.105
(d) Real Estate
TABLE 4A-D. AVERAGE CDS PER SUSTAINALYTICS ESG QUINTILES WITHIN SPECIFIC SECTORS
However, these aggregated metrics might lead to overestimation of the true relationships within. Figure
1 presents the data universe in ESG/CDS space. Any relationships are almost completely obscured.
Even the Information Technology provides little confidence in the relationship.
5
(a) Normal scale
PHASE II
OLS linear regression was run for the CDS on the ESG scores. Since CDS is obviously influenced much
more heavily by other factors, we do not expect to see significance in these regressions. Subsequent
regressions iteratively add categorical variables to control for S&P credit ratings, as well as economic
sector. Of particular interest, we introduce construction variables of the products of the sectors and the
E, S, and G scores, respectively, in order to examine if E, S, or G scores have a disproportionate effects in
particular sectors. For example, it might be reasonable that Environmental Scores would have more
6
predictive strength on CDS within the Energy or Materials sectors, whereas Social Scores might have
more predictive strength in the Health Care sector.
CDS Regressed Over E, S, and G, Credit Ratings, Sector, and Sector/ESG Interaction
Dependent variable: Credit Default Swap
7
and Industrials Sector because it contained the largest number of data points.
Some coefficients for Telecommunication Services not shown because there were only two investment
grade Telecommunication Services companies in the S&P500 with CDS data.
Before discussion of the regressions in Table 5, it is important to note the adjusted R2 values.
Essentially, the R2 value represents the percentage of the variation in the output (CDS in this case) that
can be reasonably explained by the particular model. It must be adjusted given that fitting over more
variables necessarily increases the R2 value regardless of whether or not the added variables actually
have any predictive power, and ultimately over-fitting the data. Therefore, even for model (3), which
regresses over E, S, G, the S&P Credit Rating, and Sector, is only able to explain about 31.7% of the
variation exhibited by the CDS. That having been said, the models may give insight to relative
importance within ESG considerations as opposed to pure predictive power.
The disappearance of the significance of the Sustainalytics Social Score after the first two models
suggests that the CDS is better explained by sector level factors than aggregate company factors. The
drop in decrease in adjusted R2 value from model (3) to model (4) seems to suggest that, given the size
of the data universe, no more useable information is able to be extracted by examining E, S, and G
influence at a sector-specific level. Only Governance in the Consumer Staples showed any statistical
interaction, with higher Governance scores explaining a slight rise in CDS.
8
(b) Logarithmic Y-Scale
Due to the lack of financial data, S&P credit ratings may be used as a proxy for a holistic view of the
financial situation of each company. Figure 2 shows the expected result of decreasing, and generally
less variable, CDS with higher credit ratings. This can also be seen in Table 5, albeit without statistical
relevance other than the BBB- rating.
The correlation between the ESG, E, S, and G scores of Sustainalytics and MSCI is examined.
9
FIGURE 3. SUSTAINALYTICS AND MSCI ESG SCORES FOR S&P500
Table 6 and Figure 3 illustrate the low correlation between the respective ESG, E, S, and G scores of
Sustainalytics and MSCI. Even though, the calculation methodology between these two research
companies differs, one might still have expected a higher degree of correlation between their scores on
the same companies. That having been said, perfect correlation between the two would eliminate the
need for one or the other. The especially weak correlation in Governance scores points to a possible
lack of governance related data or structure around interpretation of Governance practice.
CDS is regressed on the difference between the two scaled ESG scores as well as S&P credit ratings, and
sectors. This is done because it might be plausible that severe differences in the ESG scores for a given
company might signal that there is a lack of reliable information, or perhaps large differences in
interpretation, about that company’s activities in those areas, and thus present risk and producing
upward pressure on the CDS.
CDS Regressed Over Absolute ESG Difference, Credit Ratings, and Sector
Dependent variable:
Credit Default Swap
(1) (2) (3)
Absolute Difference in Scaled ESG Scores -0.560 -0.194 -0.286
S&P Credit Rating: BBB- 115.672*** 98.429***
S&P Credit Rating: BBB 59.580* 45.873
S&P Credit Rating: BBB+ 46.299 34.829
S&P Credit Rating: A- 34.477 23.664
S&P Credit Rating: A 21.392 14.438
S&P Credit Rating: A+ 11.713 12.532
10
S&P Credit Rating: AA- 21.535 11.470
S&P Credit Rating: AA 4.304 12.648
S&P Credit Rating: AA+ 8.164 -31.707
GICS Sector: Consumer Discretionary 29.633**
GICS Sector: Consumer Staples 0.485
GICS Sector: Energy 44.447***
GICS Sector: Financials 17.771
GICS Sector: Health Care -5.203
GICS Sector: Information Technology 15.905
GICS Sector: Materials 8.430
GICS Sector: Real Estate 47.272***
GICS Sector: Telecommunication Services 17.259
GICS Sector: Utilities 13.799
Constant 81.150*** 32.227 28.367
(6.029) (32.744) (33.083)
Adjusted R2 0.0001 0.254 0.315
Residual Std. Error 52.009 (df = 201) 44.912 (df = 192) 43.057 (df = 182)
*
Note: p<0.1; **p<0.05; ***p<0.01
TABLE 7. CDS REGRESSED ON DIFFERENCE IN ESG SCORING, CREDIT RATING, AND SECTOR
Table 7 exhibits no significant relationship between CDS reacting to a relative difference in Sustainalytics
and MSCI ESG scoring. Arguably, the lack of significant influence of ESG scores on CDS promotes a lack
of interest in the difference of such scores in the CDS market, making the difference in said scores
effectively meaningless.
DISCUSSION
Previous literature has concluded that ESG scores indicate small, but positive, pressure on corporate
bond performance (Clubb et al., 2016; Desclée et al., 2016). For the purposes of evaluating the
investment grade companies within the S&P500, the further delineation of the sectors revealed certain
sectors that seems to exhibit higher correlations between ESG scores and CDS. However, the limited
width and depth of the universe precluded further granularity for examining statistically relevant E, S,
and G specific relationships within given sectors. Not enough financial data was available to control for
more relevant factors, and the weakness of the resulting R2 values dampens these findings.
Nevertheless, the relationships exposed by the regression, rather than its predictive use, are the
relevant outcomes. It should also be noted that ordinary least squares regression relies on several
assumptions, including that the residuals are normally distributed and homoskedastic, the errors are
independent, and the relationships are linear. Further research could address the control over a greater
number of financial variables using high-dimensional regression, as well as adjusting the linear model for
robustness against heteroskedasticity.
11
As the relative ESG scores of Sustainalytics and MSCI were found to be very weakly correlated, a
regression of CDS on the difference between the scaled ESG scores, S&P credit ratings, and economic
sector produced little inference between scaled ESG scores predicting a reduction in CDS. This points to
the need for standardization in ESG reporting and the policy space around corporate disclosure of
sustainability measures.
PHASE III
In Phase III, three cases in the sectors with positive correlations between ESG scores and the CDS
spreads were selected based on the results of Phase I. The Industrial, Material and Technology sectors
showed positive correlations between the ESG scores and the CDS spreads. Then, we matched the ESG
scores with the controversy report.
FINANCIALS
GE has a revenue of $122B, profit of $1.68B, margin of 1.39%, stock share of $28.8 and market Cap of
$285.6B. The Sustainalytics ESG score in Q4 2016 was 67.67 and the CDS 47.30 BP, in Q1 2017 the
Sustainalytics ESG score is 70.16 and the CDS Spread of 37.86 BP. “Even though credit default swaps
(CDS) are basically insurance policies against the default of a bond issuer, many investors had used these
securities to take a view on a particular credit event” (Brown, 2014). In this particular case, an
improvement in the ESG score could indicate a correlation with the drop in CDS spread.
Quintile 1 2 3 4 5
ESG 71.13143 62.53429 52.145 37.77286 50.28833
E 80.68714 55.95143 40.83333 46.91286 47.46833
S 56.25714 66.66714 60.435 41.63 50.03
G 69.18857 52.48571 57.165 56.83429 37.02
12
We wanted to see if there is a relationship between the ESG scores and the Controversy scores to see if
the ESG scores can show signals before any controversy or incident occurs. In order to see the
relationship, we tracked the changes of ESG scores and matched them with controversy cases.
Unfortunately, the given controversy scores were the recent 3~5 years data while the ESG scores started
from 2008. With the purpose of delivering clear visualization, the controversy scores indicate the
severeness of controversies, and the score 10 means “Very Severe” while the lower score indicates
“Minor”, as following:
GE had quite steady ESG scores over times in the range of 55 to 78, but recently, the Social score and
the Environment score went down while the Governance score stayed high. Most controversies in GE
were related to Social issues including labor management, health & safety, product liability, and impact
on local communities. When the Social score in ESG started to reduce, a severe controversy related to
labor management occurred. On November 10th, 2015, there was a severe social controversy that many
labor unions filed a class action lawsuit against GE because the company terminate medicare plans and
drug benefits for retired salaried workers, affecting approximately 130,000 retirees and their spouses.
Likewise, right after Environment score decreased in late 2011, a severe controversy related to toxic
emissions and waste occurred. On the General Electric Company 2011 annual report, GE reported that
the cleanup of the Hudson River, the river polluted by dispensing polychlorinated biphenyls from GE,
may take longer time than estimated because the contamination was severe in wider areas. With the
chart with GE’s ESG scores and Controversy scores, we could see some relationships between ESG
scores with very severe controversies.
13
Figure 4. ESG Scores and Controversy Scores for General Electric Company
FINANCIALS
Apple Inc. has a revenue of $233B, profit of $53.7B, margin of 23%, Stock share of $153 and market Cap
of $586B. The Sustainalytics ESG score in Q4 2016 was 67.36 and the CDS 47.30 BP, in Q1 2017 the
Sustainalytics ESG score is 68.28 and the CDS Spread 37.86 BP. As mentioned above credit default swaps
(CDS) is comparable to an insurance policies against a unreliable a bond issuer. In this particular case, an
improvement in the ESG score could indicate a correlation with the drop in CDS spread.
Quintile 1 2 3 4 5
ESG 106.9367 102.4633 82.35667 70.33333 67.555
E 106.9367 102.4633 64.75667 83.69333 73.915
S 106.9367 110.5367 74.28333 70.33333 67.555
G 102.6867 73.78667 103.26 82.35667 67.555
14
CONTROVERSY SCORES WITH ESG
Apple Inc.’s ESG score shows its divergent approach to Environment,Social, and Governance strategy.
The company’s Environment score is quite high in between 65 to 80 while the Social score is between 35
to 55. Although it is hard to find the direct relationship that ESG scores help to predict controversies
specifically, the number of controversies is reflected by the ranges of E, S, and G scores. With the given
controversy scores, only two out of 25 controversies were related to environment while 15 of them
were social controversies. A severe environmental controversy was about environmental impacts of
Apple suppliers in China. The Institute of Public and Environmental Affairs (IPE), a Chinese
environmental NGO, reported the several years’ information about Apple suppliers’ pollution impacts
including toxic wastewater, hazardous chemicals in China in 2011. The Apple’s bad supply chain
management affected not only the environmental impacts but also social impacts. Apple’s social
controversies brought notorious issues regarding its labor management relations, child labors, and
supply chain labor standards, and ruined the company’s image significantly. Several reports showed that
Apple’s supply chain such as Foxconn and Pegatron in China hired workers under the Chinese labor laws’
minimum working age of 16, with the same number of working hours and under the same intense
working conditions as the adult workers. In addition, some new reports broadcasted the workers in its
supply chain in China who were committed suicide or diagnosed with leukemia due to the poor working
conditions and the exposure to chemicals in early 2010s. The low Social scores in ESG system connected
to the frequent controversies related to social issues.
With the given data about controversy scores, the event which was repeated several times over
different periods was counted as one controversy. In order to show the continuity and severity of
controversies, we added a chart per each case with the repeated controversies, counting separately. The
following chart is the Apple’s case as an example, and the other cases’ charts will be found in the
Appendix D.
15
Figure 6. ESG Scores and Controversy Scores for Apple Inc. - Counting Repeated Controversies
FINANCIALS
E.I. Du Pont has a revenue of $57.9B, profit of $1.95B, margin of 3.36%, stock share of $79.9 and market
Cap of $57.5. The Sustainalytics ESG score in Q4 2016 was 58.55 and the CDS 55.10 BP, in Q1 2017 the
Sustainalytics ESG score is 59.57 and the CDS Spread 53.03 BP. In this particular case, an improvement
in the ESG score could indicate a possible correlation with the drop in CDS spread.
Quintile 1 2 3 4 5
ESG 74.91 76.96 72.99 83.19 52.57
E 64.62667 87.24333 93.5 50.59 64.66
S 74.91 73.81 76.14 77.04667 58.71333
G 83.5 72.86333 58.87333 77.48333 67.9
16
While the DuPont’s total ESG scores over the periods didn’t change much, the Social scores increased
and the Environment scores decreased. The latest Social and Environment scores dramatically changed:
the Social score on January, 2017 went up to 70.29 while the Environment score in the same period
dropped to 47.67. On the other hand, the Governance score stayed stable above 65. As the Environment
scores and Social scores are relatively low, there are more controversies related to environmental issues
and social issues. Since DuPont is a chemical company, many environmental controversies were related
to toxic emissions and waste and its impacts on local communities, biodiversity, and land use. The most
severe environmental controversy showed that many local and international communities and NGOs
have criticized that the DuPont’s continued development of genetically modified (“GM”) crops damaged
the biodiversity and their contamination harmed other local crops since 2012. In addition, in 2016, the
company faced many lawsuits regarding the groundwater contamination by the leakage of
perfluorooctanoate acid which can cause “kidney cancer, testicular cancer, ulcerative colitis, thyroid
disease, hypercholesterolemia (high cholesterol), and pregnancy-induced hypertension”
(Perfluorooctanoic acid, 2017). DuPont also had a very severe social issue that four workers were killed
in gas leak in Texas plant, so the U.S. Occupational Safety and Health Administration charged DuPont
with penalties for the serious and repeat violations against its employees’ health and safety. The low
Environment and Social score in ESG brought more controversies in environmental and social issues.
Figure 7. ESG Scores and Controversy Scores for E. I. Du Pont De Nemours and Company
The three cases with GE, Apple, and DuPont shows that the number of controversies is affected by ESG
scores. If the company has the higher Environment score, there are the fewer number of controversies
related to environmental issues. However, it was hard to find the direct relationship between a ESG
score and a controversy score. Therefore, we can’t predict specific controversies by certain ESG scores,
but we can assume that the higher ESG scores can bring fewer negative controversies.
17
REFERENCES
Brown, M. (2014, September 14). Credit Default Swaps: What Happens In A Credit Event? Retrieved from
Investopedia: http://www.investopedia.com/articles/bonds/09/what-happens-to-single-name-
cds.asp
Clubb, R., Takahashi, Y., & Tiburzio, P. (2016). Evaluating the Relationship Between ESG and Corporate
Fixed Income. Retrieved from http://mitsloan.mit.edu/actionlearning/labs/s-lab-projects.php
Desclée, A., Dynkin, L., Hyman, J., & Polbennikov, S. (2016). Sustainable investing and bond returns.
Retrieved from
https://www.investmentbank.barclays.com/content/dam/barclaysmicrosites/ibpublic/
documents/our-insights/esg/barclays-sustainable-investing-and-bond-returns-3.6mb.pdf
Forbes. (2016, May). Retrieved from #78 EI du Pont de Nemours:
https://www.forbes.com/companies/ei-du-pont-de-nemours/
Hlavac, Marek (2015). stargazer: Well-Formatted Regression and Summary Statistics Tables. R package
version 5.2. http://CRAN.R-project.org/package=stargazer
Perfluorooctanoic acid. (2017). Retrieved from Wikipedia:
https://en.wikipedia.org/wiki/Perfluorooctanoic_acid
18
APPENDIX
Appendix A: Quintile Tables
Note: Quintile 5 represents the set of companies with the highest ESG, E, S, and G scores, respectively.
Green: Lower CDS | Red: Higher CDS
All Sectors
Quintile 1 2 3 4 5
ESG 95.57146 65.69244 78.7355 76.71439 64.628
E 95.94659 79.6739 75.742 61.29878 68.707
S 91.09976 69.91659 74.58 71.04951 74.84375
G 76.0439 84.53195 80.188 71.85244 68.86425
Quintile 1 2 3 4 5
ESG 162.91 47.108 75.83333 80.464 126.51
E 162.91 47.384 96.67667 75.904 105.782
S 136.875 58.454 95.18167 127.51 76.142
G 102.45 53.34 114.435 142.338 84.634
Consumer Staples: 19
Quintile 1 2 3 4 5
ESG 53.2475 54.4525 80.65 49.745 47.48
E 102.0675 51.615 44.505 40.7725 46.32667
S 59.2575 48.4425 96.09 34.305 47.48
G 44.375 41.4975 47.3775 54.0475 115.21
Energy: 17
Quintile 1 2 3 4 5
ESG 149.275 89.05667 98.61 107.8367 96.16667
E 103.065 152.4633 97.265 107.8367 96.16667
S 144.53 105.9033 91.575 91.31333 111.55
G 82.6025 159.35 97.1325 128.4667 96.11
Financials: 24
Quintile 1 2 3 4 5
19
ESG 83.944 65.526 78.25 61.022 66.1675
E 69.102 77.2 86.924 45.952 78.1225
S 81.14 45.168 98.022 56.356 76.2375
G 73.568 88.772 65.546 71.41 52.975
Healthcare: 21
Quintile 1 2 3 4 5
ESG 71.02 46.1325 45.6925 54.7475 23.2925
E 60.6 56.1275 38.03 42.4025 46.33
S 58.212 57.6875 63.365 41.53 23.2925
G 67.58 58.685 51.8975 30.4425 33.14
Industrials: 33
Quintile 1 2 3 4 5
ESG 71.13143 62.53429 52.145 37.77286 50.28833
E 80.68714 55.95143 40.83333 46.91286 47.46833
S 56.25714 66.66714 60.435 41.63 50.03
G 69.18857 52.48571 57.165 56.83429 37.02
Information Technology: 14
Quintile 1 2 3 4 5
ESG 106.9367 102.4633 82.35667 70.33333 67.555
E 106.9367 102.4633 64.75667 83.69333 73.915
S 106.9367 110.5367 74.28333 70.33333 67.555
G 102.6867 73.78667 103.26 82.35667 67.555
Materials: 15
Quintile 1 2 3 4 5
ESG 74.91 76.96 72.99 83.19 52.57
E 64.62667 87.24333 93.5 50.59 64.66
S 74.91 73.81 76.14 77.04667 58.71333
G 83.5 72.86333 58.87333 77.48333 67.9
Real Estate: 13
Quintile 1 2 3 4 5
ESG 87.99333 100.2667 135.055 130.6933 127.045
E 87.99333 123.2167 83.81 141.9067 127.045
20
S 97.78667 160.4967 93.58 118.1233 82.34
G 87.99333 118.3233 121.625 142.8833 95.105
Utilities: 18
Quintile 1 2 3 4 5
ESG 66.4625 83.31 109.5233 54.045 61.06
E 57.9725 58.415 160.62 49.1075 61.06
S 59.9775 89.8675 63.83667 92.425 55.47667
G 54.72 94.865 40.57333 112.185 52.74
Appendix B: CDS Regressed Over E, S, and G, Credit Ratings, Sector, and Sector/ESG Interaction
Dependent variable:
Credit Default Swap
(1) (2) (3) (4)
Sustainalytics Environmental Score -0.792** -0.568* -0.393 0.164
(0.322) (0.291) (0.335) (1.009)
Sustainalytics Social Score -0.039 0.260 -0.125 0.316
(0.395) (0.349) (0.423) (1.107)
Sustainalytics Governance Score -0.155 -0.451 -0.102 -1.477
(0.459) (0.405) (0.520) (1.345)
S&P Credit Rating: BBB- 102.992*** 91.401*** 107.116***
(33.628) (33.010) (37.512)
S&P Credit Rating: BBB 44.345 37.542 46.232
(32.998) (32.565) (36.244)
S&P Credit Rating: BBB+ 33.454 28.413 37.309
(32.733) (32.341) (36.081)
S&P Credit Rating: A- 20.449 16.942 24.364
(33.129) (32.754) (36.663)
S&P Credit Rating: A 6.896 7.846 14.661
(32.990) (32.649) (36.616)
S&P Credit Rating: A+ 2.715 7.274 25.232
(35.114) (34.482) (37.744)
S&P Credit Rating: AA- 9.546 7.702 22.936
(34.565) (34.291) (38.116)
S&P Credit Rating: AA -3.657 7.048 47.027
(40.771) (40.116) (45.809)
S&P Credit Rating: AA+ -12.617 -36.196 -35.647
(55.120) (54.687) (60.958)
GICS Sector: Consumer Discretionary 28.236** 136.350
21
(11.602) (146.415)
GICS Sector: Consumer Staples 2.835 -215.763
(12.951) (136.725)
***
GICS Sector: Energy 42.704 117.884
(14.726) (181.781)
GICS Sector: Financials 15.120 -2.462
(13.730) (148.782)
GICS Sector: Health Care -4.071 -104.014
(12.950) (134.752)
GICS Sector: Information Technology 20.064 34.590
(14.719) (181.025)
GICS Sector: Materials 10.498 -12.310
(14.155) (148.146)
GICS Sector: Real Estate 47.996*** -32.652
(14.750) (158.117)
GICS Sector: Telecommunication Services 20.596 191.991
(32.134) (363.760)
GICS Sector: Utilities 13.980 127.926
(14.548) (202.009)
Environmental Score: GICS Sector: Consumer
-1.759
Discretionary
(1.377)
Environmental Score | GICS Sector: Consumer
-2.375
Staples
(1.634)
Environmental Score | GICS Sector: Energy 0.854
(3.012)
Environmental Score | GICS Sector: Financials -0.437
(1.300)
Environmental Score | GICS Sector: Health Care -0.302
(1.509)
Environmental Score | GICS Sector: Information
0.934
Technology
(1.855)
Environmental Score | GICS Sector: Materials -1.105
(1.758)
Environmental Score | GICS Sector: Real Estate 0.454
(2.313)
Environmental Score | GICS Sector:
-2.349
Telecommunication Services
(5.048)
Environmental Score | GICS Sector: Utilities -0.085
(1.511)
22
Social Score | GICS Sector: Consumer Discretionary -1.571
(1.706)
Social Score | GICS Sector: Consumer Staples 1.483
(1.880)
Social Score | GICS Sector: Energy -1.564
(2.400)
Social Score | GICS Sector: Financials -0.255
(2.260)
Social Score | GICS Sector: Health Care -1.722
(1.886)
Social Score | GICS Sector: Information Technology -2.283
(2.316)
Social Score | GICS Sector: Materials -0.122
(1.940)
Social Score | GICS Sector: Real Estate -0.235
(1.690)
Social Score | GICS Sector: Telecommunication
Services
Social Score | GICS Sector: Utilities 0.690
(1.964)
Governance Score | GICS Sector: Consumer
1.222
Discretionary
(2.028)
Governance Score | GICS Sector: Consumer Staples 4.055*
(2.102)
Governance Score | GICS Sector: Energy -0.445
(3.101)
Governance Score | GICS Sector: Financials 0.781
(2.107)
Governance Score | GICS Sector: Health Care 3.109
(2.133)
Governance Score | GICS Sector: Information
0.826
Technology
(2.985)
Governance Score | GICS Sector: Materials 1.418
(2.189)
Governance Score | GICS Sector: Real Estate 0.909
(3.321)
Governance Score | GICS Sector:
Telecommunication Services
Governance Score | GICS Sector: Utilities -1.910
(2.971)
Constant 137.686*** 93.690** 70.726 95.689
23
(32.439) (45.114) (49.080) (97.579)
Adjusted R2 0.026 0.271 0.317 0.267
51.337 (df = 44.397 (df = 42.983 (df = 44.520 (df =
Residual Std. Error
199) 190) 180) 152)
*
Note: p<0.1; **p<0.05; ***p<0.01
Appendix C: CDS Regressed Over Absolute ESG Difference, Credit Ratings, and Sector
Dependent variable:
Credit Default Swap
(1) (2) (3)
Absolute Difference in Scaled ESG Scores -0.560 -0.194 -0.286
(0.556) (0.492) (0.484)
S&P Credit Rating: BBB- 115.672*** 98.429***
(33.575) (32.862)
S&P Credit Rating: BBB 59.580* 45.873
(32.676) (32.209)
S&P Credit Rating: BBB+ 46.299 34.829
(32.687) (32.246)
S&P Credit Rating: A- 34.477 23.664
(32.994) (32.583)
S&P Credit Rating: A 21.392 14.438
(32.830) (32.432)
S&P Credit Rating: A+ 11.713 12.532
(35.220) (34.355)
S&P Credit Rating: AA- 21.535 11.470
(34.817) (34.204)
S&P Credit Rating: AA 4.304 12.648
(41.034) (39.981)
S&P Credit Rating: AA+ 8.164 -31.707
(55.163) (54.646)
GICS Sector: Consumer Discretionary 29.633**
(11.589)
GICS Sector: Consumer Staples 0.485
(12.730)
GICS Sector: Energy 44.447***
(13.654)
GICS Sector: Financials 17.771
(11.773)
GICS Sector: Health Care -5.203
24
(12.558)
GICS Sector: Information Technology 15.905
(14.286)
GICS Sector: Materials 8.430
(13.800)
GICS Sector: Real Estate 47.272***
(14.448)
GICS Sector: Telecommunication Services 17.259
(32.057)
GICS Sector: Utilities 13.799
(13.172)
***
Constant 81.150 32.227 28.367
(6.029) (32.744) (33.083)
Adjusted R2 0.0001 0.254 0.315
Residual Std. Error 52.009 (df = 201) 44.912 (df = 192) 43.057 (df = 182)
*
Note: p<0.1; **p<0.05; ***p<0.01
Appendix D: Controversy Scores and ESG Scores of GE and DuPont – Counting repeated controversies
25
26