Publications

Publications



COGEM provides estimates of Greenhouse Gas (GHG) emissions for a vast dataset of over 47 000 listed companies. The emission data are categorized into scopes 1, 2, 3 and 123 (sum of all scopes) for each company. Data is available from 2005 to 2024. With over 650 000 samples, it is, to our knowledge, the largest dataset of its type. It is produced by Machine Learning algorithms, of which the full methodology is detailed in the following paper link and the code can be found in the public Github repository.

- Barreau, T., Fahmaoui, M., Tankov, T., SSRN (2024).



Climate Finance Bench introduces an open benchmark that targets question-answering over corporate climate disclosures using Large Language Models. We curate 33 recent sustainability reports in English drawn from companies across all 11 GICS sectors and annotate 330 expert-validated question-answer pairs that span pure extraction, numerical reasoning, and logical reasoning. Building on this dataset, we propose a comparison of RAG (retrieval-augmented generation) approaches. We show that the retriever's ability to locate passages that actually contain the answer is the chief performance bottleneck. We further argue for transparent carbon reporting in AI-for-climate applications, highlighting advantages of techniques such as Weight Quantization. The code and the associated data can be found in the public Github repository.

- Mankour, R., Chafai, Y., Saleh, H., Hassine, G. B., Barreau, T., & Tankov, P., ArXiV (2025).



We develop a new bottom-up methodology to estimate companies' (mis)alignment to net-zero scenarios. It uses companies' asset-level data on greenhouse gas emissions at production units. We apply the methodology to the steel sector globally and we find that companies' projected emissions at 2030 exceed by up to 20% the levels of the corresponding net-zero scenario of the International Energy Agency, depending on the rate of decarbonization of electricity supply to steel production. Further, we find that projected emissions at 2030 exceed companies’ aggregate stated targets, even in the optimistic case of electricity supply decarbonization rate following the net-zero scenario. Moreover, the discrepancy is driven by the largest steel companies. Our results show that a bottom-up asset-level approach allows for a reality check of companies’ contributions to national decarbonization plans. This, in turn, is crucial to inform more targeted industrial policies for decarbonization, and regulatory disclosure.

- Saleh, H., Battiston, S., Monasterolo, I., Barreau, T., & Tankov, P., SSRN (2025).



In this paper, we present a critical raw materials index (CRMI) that represents the price dynamics of the raw materials required for the low-carbon transition. Using a unique market and trade dataset covering 29 critical raw materials from 2012 to 2023, we construct a weekly trade weighted price index following a robust methodological framework. The relevance of our index is demonstrated through a validation process including a plausibility analysis and a comparability analysis. In addition, a sensitivity analysis provides empirical evidence of the robustness of our index to alternative data treatment, weighting factors and weighting schemes. Our framework offers policymakers a useful price benchmark to track the underlying metal market dynamics required by the growing clean energy sectors.

- Hasse, J. B., & Nobletz, C., AMSE working paper, (2024).



The economic transformation required to reach global net zero goals relies on the mining and transformation of certain minerals and metals for the production of low-carbon technologies. Increasing global demand for these critical materials, combined with their uneven geographical distribution, raise potential supply issues that pose economic security and transition risks for the European Union. This report sets out the nature of the challenge, the policy responses and ways forward for the bloc.

- Nobletz, C., Svartzman, R., Dikau, S., CETEX, (2024).