International climate finance is key to achieving the goals of the Paris Agreement. Here we develop a machine learning classifier to identify international climate finance from 2.7 million official development assistance projects between 2000 and 2019, resulting in a consistent and replicable inventory of 82,023 bilateral climate finance projects (US$80 billion). Our findings reinforce concerns that the actual numbers may be much lower than current estimates made with Rio markers.
This is a preview of subscription content, access via your institution
Subscribe to Nature+
Get immediate online access to the entire Nature family of 50+ journals
Subscribe to Journal
Get full journal access for 1 year
only $8.25 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Tax calculation will be finalised during checkout.
Get time limited or full article access on ReadCube.
All prices are NET prices.
The underlying data for the analysis in this study and the annotated training set for the language classifiers can be retrieved from https://github.com/MalteToetzke/consistent-and-replicable-estimation-of-bilateral-climate-finance. For access to the raw data and the trained model parameters of ClimateFinanceBERT, please contact M.T. (email@example.com).
All scripts used for pre-processing, modelling and analysis are publicly available via https://github.com/MalteToetzke/consistent-and-replicable-estimation-of-bilateral-climate-finance.
Burke, M., Hsiang, S. M. & Miguel, E. Global non-linear effect of temperature on economic production. Nature 527, 235–239 (2015).
Carleton, T. A. & Hsiang, S. M. Social and economic impacts of climate. Science 353, aad9837 (2016).
Climate Finance Provided and Mobilised by Developed Countries: Aggregate Trends Updated with 2019 Data (OECD, 2021); https://doi.org/10.1787/5F1F4182-EN
Rio Markers for Climate Handbook (OECD, 2016).
Yeo, S. Where climate cash is flowing and why it’s not enough. Nature 573, 328–331 (2019).
Climate Finance Shadow Report 2020 (Oxfam International, 2020).
Halimanjaya, A. Climate mitigation finance across developing countries: what are the major determinants? Clim. Policy 15, 223–252 (2015).
Weikmans, R. & Roberts, J. T. The international climate finance accounting muddle: is there hope on the horizon? Clim. Dev. 11, 97–111 (2017).
Michaelowa, A. & Michaelowa, K. Coding error or statistical embellishment? The political economy of reporting climate aid. World Dev. 39, 2010–2020 (2011).
Donner, S. D., Kandlikar, M. & Webber, S. Measuring and tracking the flow of climate change adaptation aid to the developing world. Environ. Res. Lett. 11, 054006 (2016).
Roberts, J. T. et al. Rebooting a failed promise of climate finance. Nat. Clim. Change 11, 180–182 (2021).
Climate Adaptation Marker: Quality Review (OECD, 2013).
Weikmans, R., Roberts, J. T., Baum, J., Bustos, M. C. & Durand, A. Assessing the credibility of how climate adaptation aid projects are categorised. Dev. Pract. 27, 458–471 (2017).
Joint Report on Multilateral Development Banks’ Climate Finance (AfDB et al., 2021).
Toetzke, M., Banholzer, N. & Feuerriegel, S. Monitoring global development aid with machine learning. Nat. Sustain. https://doi.org/10.1038/s41893-022-00874-z (2022).
Climate Finance in 2013–14 and the USD 100 Billion Goal (OECD & CPI, 2015); http://www.oecd-ilibrary.org/environment/climate-finance-in-2013-14-and-the-usd-100-billion-goal_9789264249424-enhttps://doi.org/10.1787/9789264249424-en
Egli, F. & Stünzi, A. A dynamic climate finance allocation mechanism reflecting the Paris Agreement. Environ. Res. Lett. 14, 114024 (2019).
Timperley, J. The broken $100-billion promise of climate finance—and how to fix it. Nature 598, 400–402 (2021).
Scott, S. The Grant Element Method of Measuring the Concessionality of Loans and Debt Relief Working Paper No. 339 (OECD Development Centre, 2017).
van Oldenborgh, G. J. et al. Attribution of extreme rainfall from Hurricane Harvey, August 2017. Environ. Res. Lett. 12, 124009 (2017).
Otto, F. E. L. et al. Anthropogenic influence on the drivers of the Western Cape drought 2015–2017. Environ. Res. Lett. 13, 124010 (2018).
Devlin, J., Chang, M. W., Lee, K. & Toutanova, K. BERT: pre-training of deep bidirectional transformers for language understanding. Preprint at arXiv (2018).
Webersinke, N. et al. ClimateBert: a pretrained language model for climate-related text. Preprint at arXiv (2021).
Sanh, V., Debut, L., Chaumond, J. & Wolf, T. DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter. Preprint at arXiv (2019).
Montani, I. & Honnibal, M. Prodigy: a new annotation tool for radically efficient machine teaching (2017). https://explosion.ai/blog/prodigy-annotation-tool-active-learning Accessed: 2022.08.15
We thank seminar participants at ETH Zurich and the University of St. Gallen and staff of the Swiss Federal Office for the Environment for valuable input on previous versions of this paper. We also thank R. Weikmans and J. T. Roberts for making available the project-level data used in Weikmans et al.13 and the research assistants and students for participating in the validation surveys. This project was partially funded by the Swiss Network for International Studies.
The authors declare no competing interests.
Peer review information
Nature Climate Change thanks Axel Michaelowa, J. Roberts and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended Data Fig. 1 Bilateral climate finance in ODA over time including projects with focus on environment.
Bilateral climate finance in ODA over time including projects with focus on environment.
Precision scores of Rio marker reporting on project-level by contributing country. The precision score is measured as the share of projects tagged as climate finance through the Rio markers that are also classified as climate finance by ClimateFinanceBERT. For the analysis, only projects tagged with principal Rio markers were considered. NaN values represent cases where no corresponding finance has been reported via the Rio markers with a principal marker. The period of evaluation is from 2016 to 2019.
Recall scores of Rio marker reporting on project-level by contributing country. The recall is measured as share of projects classified as climate finance by ClimateFinanceBERT that are also tagged as climate finance through the Rio markers. For the analysis, projects tagged with principal and significant Rio markers were considered. NaN values represent cases where no corresponding finance has been classified by ClimateFinanceBERT. The period of evaluation is from 2016 to 2019.
Financial instruments used by contributors for mitigation and adaptation between 2016 and 2019. The use of loans in ICF is concentrated among Germany, France, Japan, and Canada. Hence, one would expect absolute changes for these countries’ ICF if grant-equivalents are used.
Robustness check for mitigation and adaptation finance in ODA by recipient, calculating a grant-equivalent of 50% per loan for 2016–2019. The calculation of grant-equivalents for loans is required for data since 2018 but has only been applied to 18% of the data since then. The average calculated grant-equivalent per loan is 58%. Y-axis shows that share of the total adaptation (square) and mitigation (dot) finance that a country received since the Paris Agreement (cumulated, 2016–2019). X-axis shows the national GDP in 2019. Countries that received more mitigation finance than adaptation finance are coloured in green. Countries that received more adaptation finance than mitigation finance are coloured in orange.
Robustness check for mitigation and adaptation finance in ODA by contributor, calculating a grant-equivalent of 50% per loan for 2016–2019. The calculation of grant-equivalents for loans is required for data since 2018, but has only been applied to 18% of the data since then. The average calculated grant-equivalent per loan is 58%. Note: Y-axis shows that share of the total adaptation (square) and mitigation (dot) finance that a country contributed since the Paris Agreement (cumulated, 2016–2019). X-axis shows the national GDP in 2019. Countries that contributed more mitigation finance than adaptation finance are coloured in green. Countries that contributed more adaptation finance than mitigation finance are coloured in orange.
Schematic overview of the model framework for ClimateFinanceBERT. Project descriptions that are annotated as relevant (Relevance = 1) relate to mitigation, adaptation, or environmental activities.
Results from the user study evaluating project descriptions regarding their relevance for climate finance. X-Axis indicates the share of project descriptions that have been evaluated as climate finance by the respondents (0 = no climate finance; 1 = climate finance). Coloured triangles represent the average evaluation of each respondent. The dashed lines represent the average across all respondents for each case. Every colour refers to a specific respondent and each respondent evaluated project descriptions from each case.
About this article
Cite this article
Toetzke, M., Stünzi, A. & Egli, F. Consistent and replicable estimation of bilateral climate finance. Nat. Clim. Chang. (2022). https://doi.org/10.1038/s41558-022-01482-7