Abstract
In a first paper, I use the BERTopic model to cluster development projects based on the detailed narratives from the OECD Creditor Reporting System (CRS) dataset, which contains about 5 million projects across different sectors. This method outperforms conventional classification approaches by identifying over 400 distinct topics, compared to the 234 categories in the CRS. This advanced clustering allows the discovery of hidden patterns in project distribution and development finance allocation that were previously undetected by traditional means. In a second work in progress, I am re-evaluating climate-related development finance projects and related financial flows using an adapted ClimateBERT model. Compared to OECD estimates, this approach reveals significant discrepancies in reported climate finance, suggesting a potential overestimation in official data. In addition, the research provides insights into donor behaviour and thematic trends in the allocation of climate finance. This study contributes a novel framework for classifying development finance and offers a replicable solution for monitoring financial flows, making it an essential tool for policymakers and stakeholders working to optimise aid distribution in line with global sustainability goals.
Slides available upon requests.
Pierre Beaucoral
PhD Candidate
My research interests include development economics, climate adaptation and mitigation, environment, development and climate finance.