ESG AI: Picture from Gordon Johnson from Pixabay to illustrate green or ESG AI

ESG AI: Researchpost #125

ESG AI: >10x new research on climate AI models (for banks), CO2 removal, bridge technology risks, human capital, ESG risk management and ESG bullshit, government greenium, double materiality and listed impact investing and industry versus regional diversification by Markus Leippold, Marco Wilkens, Johannes Leister, Ottmar Edenhofen, Timo Busch, Andreas Hoepner and many more (# indicates the number of SSRN downloads on April 30th, 2023)

Social and ecological research: ESG AI

Climate LLM: Enhancing Large Language Models with Climate Resources (ESG AI) by Mathias Kraus, Julia Anna Bingler, Markus Leippold, Tobias Schimanski, Chiara Colesanti Senni, Dominik Stammbach, Saeid Ashraf Vaghefi, Nicolas Webersinke as of April 17th, 2023 (#114): “Our prototype LLM agent retrieves information from general Google searches and emission data from ClimateWatch to provide reliable and accurate information. Through two exemplary experiments, we showcase how such an LLM agent can operate to enhance the accuracy and reliability of climate-related text generation. This work contributes to the exploration of LLM applications in domains where up-to-date and accurate information is critical …” (p. 6).

Complex climate scenarios: Klima-Szenarioanalysen in Banken (ESG AI) von Marco Wilkens und Johannes Leister vom April 2023: “… supervisory-motivated climate scenario analyses build on „traditional scenario analyses“ for assessing market and economic risks, but they are much more complex. This is in particular due to the need to model the interrelationship between climate data and macroeconomic data and the significantly much longer period under consideration. In addition, there is very little empirical data available for mapping climate risks, which is needed for econometric modeling of relevant relationships. Moreover, these long time periods require considerations of how banks and bank customers act over time. However, taking into account resulting dynamic bank balance sheets lead to hardly comparable results between banks. … this allows primarily a relative estimation of climate-related risks between banks than a realistic and comprehensive estimation of climate-related credit risks for individual banks. … In summary, we see climate scenario analyses as one of several important tools for transforming both the financial industry and the real economy toward the green economy“ (abstract).

Tricky CO2 removal: On the Governance of Carbon Dioxide Removal – A Public Economics Perspective by Ottmar Edenhofer, Max Franks, Matthias Kalkuhl, and Artur Runge-Metzger as of April 19th, 2023 (#25): “This paper highlights the importance of carbon dioxide removal (CDR) technologies for climate policy. We … discuss removal costs and storage duration of different technologies. … seemingly cheap removal technologies in the land sector can indeed be very expensive when increasing opportunity costs and and impermanence are appropriately accounted for. The use of non-permanent removal – though to a certain extent economically optimal – creates high liability to firms and regulators that warrants a careful and deliberative risk management“ (abstract).

Human-Climate relations: Climate Changes Affect Human Capital by Germán Caruso, Inés de Marcos, and Ilan Noy as of April 19th, 2023 (#13): “… we provide a framework for analyzing the multiple interlinkages between climate change and human capital … The framework presents two channels through which human capital is affected: direct effects on health, nutrition, and wellbeing, and indirect effects through changes in economic systems, markets, and through damage to infrastructure. … For mitigation and adaptation, we find that while these are overall clearly beneficial, they are also associated with significant human capital costs for specific sectors and groups in society. … Since there is also evidence that high human capital improves adaptation and mitigation, this suggests that adaptation and mitigation that accounts and compensates for these ‘sectoral’ losses can create a virtuous cycle that leads to positive outcomes for both climatic action and human capital“ (abstract). My comment: My fund focuses on human/social and climate topics, see Artikel 9 Fonds: Kleine Änderungen mit großen Wirkungen? – (

Responsible investment research

Risky bridges: Bridge technologies from a sustainable finance perspective by Timo Busch, Tanja Ohlson, Ana Sarantidi, and Özüm Yenen as of March 2023: “The interviewees identified risks related to a particular bridge technology, in our case LNG infrastructure, risks that stemmed from the classification of the investment into bridge technologies, and risks to the own organizations mainly in the reputational context. These risks often led to a low appetite for investing in bridge technologies. However, the asset managers also recommended that these risks could be minimized by providing a more transparent and reliable path forward for the “end of bridge” phase of the technology. In the LNG case this relates to the future utilization of hydrogen. Moreover, a classification scheme and related label for transition finance products could help increase the attractiveness of bridge technology investments, and better communication and science-based long-term decision making would help minimize risks in the context of the bridge technology” (p. 50).

Less (ESG) bullshit: Bloated Disclosures: Can ChatGPT Help Investors Process Information? (ESG AI) by Alex G. Kim, Maximilian Muhn, and Valeri V. Nikolaev as of April 27th, 2023 (#443): “By summarizing a large sample of corporate disclosures with GPT-3.5- Turbo, we show that the length of the summaries is shortened by as much as 80%, on average. Importantly, the obtained summaries appear to provide more relevant insights as compared to the underlying raw documents. Specifically, we show that summarized sentiment better explains cumulative abnormal returns around disclosure dates than raw sentiment. Building on this insight, we construct a novel and easy-to-implement measure of the degree of “bloat” in textual disclosures. … We show that bloated disclosures are associated with slower price discovery and higher information asymmetry, thus implying negative capital market consequences. Finally, we show that GPT is useful to investors interested in targeted summaries related to important topics, such as a summary of ESG-related activities” (p. 24/25).

Good E/S/G risk management: How ESG risk management can impact security risk by Miranda Carr, Yuliya Plyakha Ferenc, Blessy Varghese, Zoltán Nagy, and Guido Giese from MSCI ESG Research as of April 13th, 2023: “Our findings indicate that companies with higher E and S risk management and governance scores, and consequently higher ESG Ratings, than their peer groups had lower stock-specific risk than their peers during the 2017-2022 time period. … A key element behind this lower risk profile is … how the company itself managed these risks … our findings demonstrated that E and S risk management adds valuable informational content in portfolio management. … For social key issues, management metrics include elements such as the promotion of training and development of the workforce for companies in knowledge-intensive industries, transparency and visibility over the supply chain for companies in the retail industry, robust health and safety policies for companies in the consumer-durables sector and positive community relations for companies in the mining industry“ (p. 14). My comment: My engagement policy focuses on several of these topics see Shareholder engagement: 21 science based theses and an action plan – (

Government greenium? How Large is the Sovereign Greenium? by Sakai Ando, Chenxu Fu, Francisco Roch and Ursula Wiriadinata as of April 19th, 2023 (#18): “This paper is the first empirical study to estimate the sovereign greenium using both the twin bonds issued by Denmark and Germany, and panel regression analysis. While the estimated greenium in this paper is not large, it has been increasing over time alongside the level of sovereign green bond issuances. Whether the administrative costs associated with green bond issuance exceed the benefit is a country-specific question … It remains an open question whether the purpose of the project associated with the green bond is a key determinant of the greenium, and whether green bonds have resulted in the climate outcomes they intended to achieve” (p. 11/12).

Double materiality: Beyond Climate: ‚EU Taxonomy‘ Criteria, Materiality, and CDS Term Structure by Andreas G. F. Hoepner, Johannes Klausmann, Markus Leippold, and Jordy Rillaerts as of April 18th, 2023: “… the risks associated with water and biodiversity impacting a firm are perceived to be long-term issues, as evidenced by significantly negative effects on CDS slopes. The negative effects are weaker but still significant for pollution prevention, also suggesting a long-term vision. The financing benefits due to a firm’s commitment to pollution prevention, however, have stronger long-term implications rather than short-term advantages. In contrast, a firm’s impact on biodiversity has no such timing differential, revealing a more imminent awareness. … Overall, our findings identify the long-term focus on infrastructure firms’ financing conditions with regard to the environmental topics covered in the latest EU taxonomy beyond climate change. Moreover, they highlight the importance of considering both materiality sides, i.e., the impact of the environment on firms and the impact of firms on the environment“ (p. 23).

Listed impact? Guidance for Pursuing Impact in Listed Equities by the Global Impact Investing Network as of March 30th, 2023: “Developed with input from over 100 investors, Guidance for Pursuing Impact in Listed Equities uses the GIIN’s “Core Characteristics of Impact Investing” to provide baseline practices and expectations for asset managers seeking to achieve positive impacts in listed equities. The guidance is structured around four main aspects of listed equities impact investing: setting fund/portfolio strategy, portfolio design and selection, engagement and performance data usage. Additionally, it introduces two key concepts, investor contribution and theory of change, that investors should consider when designing and managing listed equities impact funds”.

Traditional investment research

Good industry diversification: Market Segmentation and International Diversification Across Country and Industry Portfolios by Mehmet Umutlu, Seher Gören Yargı and Adam Zaremba as of April 14th, 2023 (#22): “We conjecture that partially segmented stock indexes that are characterized by low correlation with the world market are mainly priced by local factors and should produce abnormal returns relative to a global asset-pricing model. This implies a negative relation between correlation and future index returns in the presence of segmented indexes. Empirical evidence confirms such a relationship for the sample of industry indexes, suggesting a heterogeneous segmentation. Nonetheless, we do not observe a similar pattern for country indexes. Thus, cross-industry diversification is superior to cross-country diversification. The international diversification potential of industries does not vanish during volatile periods” (abstract).


Advert for German investors: “Sponsor” my research by investing in and/or recommending my article 9 mutual fund. The fund focuses on social SDGs and midcaps, uses separate E, S and G best-in-universe minimum ratings and broad shareholder engagement. The fund typically scores very well in sustainability rankings, e.g. see this free new tool, and the performance is relatively good: FutureVest Equity Sustainable Development Goals R – DE000A2P37T6 – A2P37T