Appendix and methodology

Methodology

1. Analysis of SDG-aligned research trends
(Springer Nature / Dimensions / SN Insights)

• Data analyses research trends from 1985 to 2024, looking at the volume of SDG research, growth rates, citation counts, Altmetric scores, and full-text downloads.
• Publication metadata are drawn from Springer Nature journals indexed in Dimensions. The Dimensions SDG classification⁸ algorithm is used to identify relevant articles.
• Only research and review articles were included.
• Where analyses are limited to Springer Nature content only, this is specified, otherwise publication trends include content published by all publishers.

2. Analysis of research citations in policy
(Overton / OpenAlex / Springer Nature hybrid journal dataset)

• Policy document metadata and citation data from Overton Index.
• Documents are grouped under SDG Goals per UN taxonomy. Overton Index uses a language model to classify individual documents to SDG targets. The classifier uses the document’s title, AI-generated description, and AI-generated theme as input. The classifier first matches policy documents to individual SDG targets – the 169 specific, measurable outcomes that underpin the Goals. These are then rolled up to their parent SDG classification. Each document may have 1 or more classifications. They may also not have a classification at all, if they’re not relevant to any SDG, or don’t contain sufficient information to classify them. For this analysis, we’ve excluded documents which have insufficient metadata or insufficient machine-readable text to be classified.
• SDG classifications reflect the policy document’s focus, not the SDG relevance of the cited research. For example, a health policy document may cite social science research, which wouldn’t itself be classed as SDG 3.

• Citations were classified into two categories:
  
  Policy-to-Research: Citations to scholarly research. Overton extracts candidate references from policy documents and matches them to their metadata record using Crossref. In order to be matched, the research must have a DOI indexed in Crossref (although the DOI doesn't need to be mentioned in the citation).⁹
  
  Policy-to-Policy: Documents citing other policy documents. In order for a reference to be identified as a policy citation, the policy document must itself be indexed in the Overton database.

• Analyses were conducted by:
  
  Source type: Government, think tank, NGO, IGO, legislative body, judicial body. ¹⁰
  
  Country: Both author country of cited research (using OpenAlex author affiliation data) and policy country from the Overton database. Within Overton Index, policy documents from IGOs such as the United Nations, UNESCO, WHO, IMF, and OECD are attributed to “IGO” rather than under the country in which the headquarters are located. European Union bodies, including the Publications Office of the European Union, the Joint Research Centre, and the European Central Bank, are treated in the same way, with their outputs grouped under “EU.”  This approach recognises their international scope and avoids counting international publications as part of a country’s own policy output.
  
  Access type: OA vs closed access (using Springer Nature hybrid journal data, and OpenAlex data where noted).
  
  Journal type: Inclusive vs selective journals (based on Springer Nature classification).
  
  Content type: Article, review, letter, news item, etc. (using Springer Nature data, and OpenAlex data where noted).

• Analyses on the data:
  
  Citations:  Number of citations and number of DOIs cited across source types, policy source countries, and author countries.
  
  OA vs non-OA: Citation frequency and time to first citation compared across access types, based on a set of 11 Springer Nature hybrid journals. We compared the time of first policy citation in Overton Index to the publication date of the article as provided by Springer Nature. In a very small number of cases, we found the publication date was after the policy citation date – these cases were assumed to be data errors and were removed. We also looked separately at OA citation patterns for all publications by country using OpenAlex country and OA data.
  
  Inclusive vs selective journals: A sample of 100,000 inclusive journal articles were matched to comparators from selective journals (same year, topic, and type) and citation frequency tested across 1,000 simulation runs. Journal comparisons are based on Springer Nature titles only. Topic classification uses OpenAlex metadata, which may not fully reflect interdisciplinary or cross-topic publication patterns.
  
  Content types: Content types for publications in Nature Portfolio Research journals were provided by Springer and were matched on DOI to publications in the Overton database. We then compared policy citations across the different content types.

⁸ https://www.digital-science.com/blog/2021/04/sdg-research-classification-dimensions-ai/
⁹ https://help.overton.io/article/how-are-scholarly-references-matched-in-policy-documents/
¹⁰ https://help.overton.io/article/how-does-overton-index-classify-its-sources/

Limitations of our analysis

Coverage bias: Overton’s database draws from over 2,700 policy sources in 193 countries or territories. While Overton’s coverage is broad, it is limited by the online availability of policy documents in some regions.¹¹ For example, some countries or institutions may not publish policy outputs in accessible formats, or their materials may be behind firewalls, not machine-readable, or unavailable for scraping. As a result, certain national or subnational policymaking outputs may be underrepresented, leading to skewed insights when making global comparisons.

• China: Analysis of Chinese policy documents is particularly constrained by systemic limitations. A significant portion of Chinese policy materials is not publicly available, limiting transparency. As a result, the project focused on tracking the international use of Chinese-authored research rather than how Chinese policy cites research.
• Citation relevance: Policy citations are not a definitive indicator of research impact. Research could be cited for context, dissent, or methodology, and not necessarily because they informed a policy decision. Likewise, impact may be achieved outside of citations in policy, for example, as oral evidence or other uncited contributions. Policy citations must be viewed as just one signal among many of how research enters and influences policymaking.
• Time lag considerations: Policy citation counts may underrepresent recent research, due to the lag between publication and policy use. Likewise, citation counts for newer documents (particularly OA works) may reflect early visibility more than long-term uptake.
• OA: While OA articles are often cited more frequently in policy, the relationship is complex and potentially confounded. OA status may correlate with other factors such as topic, funder mandates, or article age. Differences in policy citation cannot be attributed to OA status alone.

¹¹ https://help.overton.io/article/how-international-are-your-sources/

SDG document coverage in Overton Index

SDG article coverage in Dimensions 2000–2024