While the topic of sustainable investment is largely addressed at the level of equity management, it is much less so when it comes to investing in bonds. And yet, the sustainability analysis of government bonds, and therefore of countries, is a subject that deserves our attention, particularly in view of the challenges that arise. BLI has developed a proprietary approach aimed at assessing the sustainability character of emerging government issuers but also at identifying the dynamics (positive and negative) in place.
At present financial markets are expressing the sustainability of an entity at three different levels: Environment (E), Social (S) and Governance (G).
Therefore, when looking at the degree of sustainability of a company or country it will often be translated as an ESG score.
As there are currently no market-defined standards for assessing the E / S and G, an institution’s sustainable score/rating from two independent researchers or asset managers are not comprehensively comparable.
From petroleum firms that develop more desirable oil refining processes to businesses that base the hiring of employees on qualifications rather than on gender, the concept of a transition to a 'sharpened’ world has spread among most sectors. Each sector must approach this transition with a different angle respective to their challenges.
The richness of asset management comes in part from the various and often different opinions, perspectives and backgrounds of people active in the industry, meaning that investing in a sustainable way is self-evident for some and irrelevant for others.
While numerous methodologies exist for assessing sustainability, each one is tailored to an interpretation of the subject and in-line with the specificities of an asset class.
The attractiveness of the equity market over the past ten years - combined with a continuous interest for active sustainable portfolio management, pushed the asset class to the forefront of sustainability assessment in the sector.
More recently, asset managers developed sustainable approaches to government bond investing.
In order to fully respond to what a sustainable active investment strategy demands, the above-mentioned government approach needs to be robust, dynamic and comprehensive.
Indeed, a solid data-based framework is needed to lay the foundation. Practitioners often use the latter to determine how sustainable an economy is. In our view, a further continuous and interpretable analysis is needed. The interpretability should bring the effort and commitment of transparency, while the dynamism should capture the momentum of information.
The following article focuses on emerging market economies. Since five to ten years, low (or negative) interest rates in developed market gave a thrust to emerging markets sovereign bonds in the asset allocation of fixed income strategy.
We will now propose a methodology to assess and interpret the continuous evolution of emerging market countries’ sustainability profiles.
In order to do so, we will briefly explain the global environment in which the analysis takes place, we will then address the methodology itself. It is composed of two distinct but complementary approaches that each respond to the solid foundation and dynamism aforementioned.
Finally, we will examine how, as active investors, we can leverage on this analysis work by putting in perspective our country allocation and confront it with the UN Sustainable Development Goals (SDGs).
The frame of analysis
The global objective of any ESG assessment is to measure the emerging humanitarian and natural risks which regulations, populations and investors consider.
Perceiving the motivation behind young entrepreneurs and taking into account those mass movements aiming to mitigate (at least) climate change and to generate more equality (around the globe) we can see that in 2020, shaping an economy with a sustainable bias can bring opportunity, investments into a country and, perhaps, in the long term, a healthier and more sustainable economy.
On the other hand, neglecting this growing awareness may bring future challenges to a country.
Among many examples, we can mention South Africa to illustrate the above.
When taking into account the World Energy Council trilemma index, the South-African energy sector isn’t well governed.
Government unresponsiveness on renewable energy projects has left the country with a major dependency on coal.
Ceteris paribus, if the South-African government had secured a minimum level of energy creation (through renewable energy), the current unstable financial position of the country would be less affected by Eskom which needs significant financial support and experiences serious production issues.
- With this in mind, we can say that understanding and comparing the current level of sustainability of a country is not an easy task. In fact, many constraints from various parts of the analysis do exist.
- While there are important datasets focusing on country-level ESG matters, such data is globally lagged. Because the aggregation of indicators at the government level takes time and is time consuming, data can carry from 2 to 10-year lag and can have a frequency updated of 2 to 4 four years. The analysis on emerging countries accentuates this effect. This lag limits our capabilities to assess the relative sustainability dynamically.
- Another constraint, independent from the availability of data itself, is to succeed in harmonising the sustainability measures across different asset classes and across asset managers without losing the sense behind it.
Considering the above, BLI has developed a proprietary and well-structured approach that aims to tackle those challenges.
The two layer approach
The objective of an ESG sovereign score is to assess the sustainability of a government, its economy and the implemented reforms.
To do so, taking into account the above-mentioned context and challenges, it is crucial to define a robust framework that includes a fundamental ESG score relying on ESG quantitative data (1st layer) augmented by a momentum score relying on a dynamic analysis of qualitative information on each country (2nd layer).
The first layer:
The first layer represents the fundamental ESG score of a country. In other words, where the country places in terms of environmental, social and governance fields comparative to others.
To achieve this, we set out to find indicators. After delving into various datasets, we narrowed our indicators probe on either aggregated indicator or raw data indicator. For the latter, we selected a specific indicator by focussing on two main components:
- The source of the data
- The last date and the frequency update of the data
The second objective of the sovereign ESG score should reflect a potential future risk, in the aggregation of the three respective E/S/G scores, we overweight the governance and equally weight the social and the environmental metrics.
The natural intuition behind the overweight of governance is related to the fact that emerging countries first need political stability as well as the establishment (and maintenance) of infrastructure (e.g. viable transportation routes, energy distribution system) in order to implement perennial social or environmental reforms.
We finally arrive at an aggregate score that represents the relative sustainability of a country.
The following scatter plot illustrates the relative position of countries on the ESG angles.
While there is a limited relationship between the social and environment angles, we find a relatively good correlation between the social and governance angle. A less clear relation results between governance and environment. This is mainly driven by oil export countries such as Saudi Arabia, UEA, Qatar, but with an above mean country governance score.
The second layer:
To have a comprehensive view of a country ESG profile (and its related risks), it is important to be able to take into account the dynamics (positive and negative) that are in place. However, as recent quantitative data are lacking, we chose to continuously and systematically analyse, for each country, qualitative information (news flow, articles…) coming from specific sources.
In order to rationalise the work - but mostly to be able to objectively compare the different news/reform/population sentiment between them we have capitalized on advances in the fields of artificial intelligence and language processing. To do so, we put in place a Universal Language Model Fine Tuned for each respective angle of ESG. (ULMFIT with RoBERTa).
Each week new qualitative ESG focussed country analysis will feed the ULMFIT and the output will be a sigmoid distribution positively or negatively skewed depending on the analyst’s view on his E and/or S and/or G sided assessment.
Through the second layer, we can continuously monitor the relative sustainability of emerging countries and see if a negative or positive sustainable move/trend of a country can translate itself into worsening/improving fundamentals on the medium term.
This method allows us to understand why a country displays a better/worse ESG score. The easy interpretability of the framework lets us build our own interpretation of the current situation of a country and take, on the medium term, an (div)-(in) - vestment decision.
Below are examples of sentences and their respective score derived from the output of the ULM Fine Tuned on the Governance angle:
- The Economist March 13, 2020: “Chile’s police force is brutal, corrupt and incompetent. They need to be thoroughly reformed.”
-> Score = 0.08
- Asian Development Bank February 11, 2020: “Today is international day of women and girls in science. We are working with Sri Lanka to boost the number of girls enrolling in stem subjects by upgrading their science programs and improving learning environments in schools.”
-> Score = 0.84
Leverage the ESG score by reflecting its impact
Finally, by comparing the country allocation of our portfolios with the framework defined by the UN-SDG, we are able to put in perspective our work as an asset allocator and assess the ESG impact of our investment decisions.
You can interpret the scores as a percentage of the accomplished journey to the 2030 UN-SDG sustainable Goal. « 100 » meaning that the target for each indicator in the framework is met.
Sources: Bertelsmann Stiftung and Sustainable Development Solutions Network (SDSN), BLI
However, this framework cannot be used as a main channel to drive an active ESG bias strategy as it suffers from the same problems of data delay or lag explained at the beginning of the article.
Nevertheless, it provides a great base for common comparison and a comprehensive way to report the impact of our country allocation.
Finally, the establishment of the ESG score has allowed us to see more clearly the material used by the industry to define a country's level of sustainability. Even though there is a vast amount of data available on different aspects of sustainability with an impressive number of ramifications, it is still too early in the transition to be able to claim a dynamic analysis based on this purely quantitative static data.
The fact that our analysis is based solely on emerging countries accentuates the sluggishness in updating and creating extra-financial data. This is partly due to the lack of incentives and/or priority for these economies. Indeed, while in the developed economies the growing weight of sustainability in the different strata of society creates inertia against environmental, social and governance unconsciousness, this resistance is still very limited in developing countries.
However, the transition has begun. Emerging economies have already been leading the way for several years on certain aspects of sustainability (e.g. Uruguay and its efforts on the environment) and are pushing others to follow suit. On the side of the asset management sector, the constant evolution of computing power is opening new doors to the analysis of the sustainability of countries. This will enable us to understand where there is room for improvement.
 Eskom is a state owned South African electricity company that provides 95% of total power generation of the country.
 The sources vary depending on the aim of the indicator. Here are some notable sources: United Nations Office for Disaster Risk Reduction / Food and health Organization / International Labour Organization, WHO-UNICEF Joint Monitoring Programme for Water Supply, Sanitation and Hygiene / WHO Global Health Observatory / World Development Indicators database, World Bank / Skills Towards Employability and Productivity (STEP) of World Bank / Trusted national statistics bureau.
 Extensive researches have been already made by well-known organizations (e.g. world governance indicators by World Bank) to preselect and aggregate indicators at an index level.
 The ULMFIT model was developed by Jeremy Howard and Sebastian Ruder. For more information please see https://arxiv.org/abs/1801.06146. The implementation was made possible thanks to the fastai python package.
 The RoBERTa model is a Facebook augmented model of BERT (google). For more information please see https://ai.facebook.com/blog/roberta-an-optimized-method-for-pretraining-self-supervised-nlp-systems/. Implemented thanks to the Hugging face- transformers python package.
 The sigmoid distribution function: fx=11+e-x. It represent a S shaped curve bound between (0,1). For more information: Weisstein, Eric W. "Sigmoid Function." From MathWorld--A Wolfram Web Resource. https://mathworld.wolfram.com/SigmoidFunction.html