Science-wide mapping and ranking of institutions based on affiliated authors

Standardized information on the size of institutions and on their concentration of high-impact scientists is missing. Here, we generate institution-level data on 6,979 institutions with >=50 authors who have had >=15 Scopus-indexed items. We present data on the number and proportion of top-cited authors (according to a composite citation indicator) for each institution. The primary analysis considers 1,214,702 authors (145,704 top-cited in career-long impact) who started publishing in 1980 or later, have published >=40 Scopus-indexed items and >=5 single-, first- or last-authored items. There is modest overlap among top institutions based on volume and based on number of top-cited authors. However, there is minimal overlap when institutions with the highest proportions of top-cited authors are considered; these include mostly research institutes and tech organizations and a minority of universities. Institutions with few eligible authors have large uncertainty. We propose a percentile ranking of the 2,380 largest institutions that adjusts the number of top-cited authors for the proportion of those who have high (>95th percentile) self-citation rates and high rates of publications in Scopus-discontinued titles and further penalizes for authorships of retracted papers not due to publisher/journal errors. Institutions with the highest self-citation, discontinuation, and retraction penalties cluster in specific countries. Among countries with >=15 institutions among the 2,380, Saudi Arabia, China, Malaysia, Iran, India and Indonesia have the lowest median percentile rankings of their institutions (3.9th-21st percentile) due to high penalties. The publicly available datasets that we provide allow institution-level research assessments balancing high impact against retractions and gaming practices.