Bibliometric Indicators, By Sustainable Development Goals (SDGs)

This dataset presents trends in the total number of scientific publications relevant to societal goals, as proxied by the United Nations Sustainable Development Goals (SDGs) framework, expressed in fractional counts, for the 2015–2023 period. Coverage includes OECD member and accession countries, key partners (Brazil, China, India, Indonesia, and South Africa), and other selected non-OECD economies.

This is an experimental indicator, developed under the OECD’s Fundstat initiative to improve evidence on the directionality and societal relevance of research and development (R&D;) activities, by linking scientific output to 17 SDGs or none. To produce this indicator, the OECD developed a probabilistic AI-assisted classification model that identifies the SDG relevance of scientific publications based on their titles and abstracts. The approach builds on insights from the OECD International Survey of Science (ISSA2021), in which researchers reported which SDG their research contributed to. Using these responses, a labelled dataset was created to fine-tune SciBERT, a large language model (LLM) pre-trained on scientific literature, enabling it to predict the probability of relevance of a given text to each of the 17 SDGs, or to none. The trained classifier is then applied to publication titles and abstracts from the Scopus Custom Data. Publication counts are fractionalised across countries and aggregated annually for comparability.

For further information on the methodology, see: Aristodemou, L., et al. (forthcoming), “Assessing the direction of R&D; funding towards the Sustainable Development Goals (SDGs) with a new classifier and the OECD Fundstat infrastructure”, OECD Science, Technology and Industry Working Papers.

Source: OECD Bibliometric Indicators, calculations based on Elsevier's Scopus Custom Data, Elsevier, Version 1.2025.

Please cite as: OECD (2025), OECD Bibliometric indicators, OECD Science, Technology and Innovation Statistics, http://oe.cd/scientometrics, accessed on xx/xx/xxxx.