The Drivers of Science Referenced in US EPA Regulatory Impact Analyses: Open Access, Professional Popularity, and Agency Involvement

We perform bibliometric analysis on documents for 255 Regulatory Impact Analyzes (RIAs) prepared by the US Environmental Protection Agency (EPA) from 1980 through 2024. Using a series of automated information extraction methods, we extract references from these documents and match them to bibliographic records. We then build a database of relevant articles (whether cited in an RIA or not) and fit a two-stage regression model that predicts whether, and how many times, a reference is used in RIAs as a function of journal prestige, professional popularity, article accessibility, EPA funding, and involvement of EPA employees as co-authors. By considering cited and uncited articles related to similar scientific concepts, we can observe systematic differences in what types of research products get used in policy analysis. Academic popularity, open access status, and EPA authorship and sponsorship all predict more likely and more frequent article use. Articles in prestigious journals are no more likely to be referenced, but once referenced in the corpus are then referenced more frequently.