New article on a consistent and replicable estimation of bilateral climate finance published in Nature Climate Change

In the paper, the authors use machine learning to identify climate-relevant projects from 2.7 million descriptions of development assistance projects and thereby provide a “consistent and replicable estimation of bilateral climate finance”. Their model classifies 80,023 projects as climate finance between 2000 and 2019, of which 52% are adaptation and 48% are mitigation projects, totaling USD 80 billion. These numbers are considerably lower than those reported by contributing countries.

The main findings in short:

- Trends and patterns: Between 2000 and 2019, bilateral climate finance flows have increased form USD 0.5 billion to 7.7 billion annually. However, after 2010, growth in climate finance has considerably slowed. 64% of the disbursements are mitigation and 35% are adaptation finance. Mitigation finance primarily flows to middle income countries with a large share of projects financed through loans. In contrast, adaptation finance is distributed more evenly across recipient countries, focusing on countries at risk, including small island developing states, and providing most disbursements as grants.

- Comparison with reported numbers: Estimates by the authors are 21% below the numbers reported as “principle” and 64% below the numbers reported as “principle or significant” climate finance. For some contributors, their machine learning classifications show clear overlaps with project tags by contributors, while, for other contributors, there is a large disparity. The authors see that many contributors not only over-report their climate finance contributions but also omit a large number of projects with a clear climate-relevance. This supports criticism on poor reporting quality and inconsistency across contributors.

At the upcoming climate conference, the COP27, climate finance will be one of the key negotiation themes. After the USD 100 billion per year target from 2020, countries will negotiate a post-2025 climate finance target. First deliberations for a post-2025 finance target have started at the COP26 and some countries have already put forward concrete numbers. However, findings from this paper suggest that, beyond the commitment of mere numbers, it is crucial that countries agree upon a definition of what counts as climate finance and establish a transparent verification process for climate finance reporting.

Their proposed machine learning model (ClimateFinanceBERT) can be replicated via Downloadhttps://github.com/MalteToetzke/consistent-and-replicable-estimation-of-bilateral-climate-financevertical_align_bottom.

For further contextualization, the climate policy researchers Robert Timmons (Brown University) and Romain Weikmans (Finish Institute of International Affairs/Université Libre de Bruxelles) have published a News&Views article on Nature Climate Change were they discuss the implications of this study for the upcoming climate negotiations at the COP27, which can be found here: Downloadhttps://www.nature.com/articles/s41558-022-01483-6vertical_align_bottom


Link to the full paper: Downloadhttps://rdcu.be/cV7sPvertical_align_bottom