How far added affiliations move university ranks depends on the counting rule — data and code

Code and aggregate data for a study of how far publications that a university holds only through authors also affiliated abroad move it in a bibliometric ranking, and how the counting rule decides how far. The study uses the OpenAlex snapshot of 25 June 2026 (CC0) and the CWTS Leiden Ranking Open Edition 2025. For 20,492 universities and every year from 2010 to 2025 it measures the share of a university's works held only through dual-affiliated authors, detects episodes in which that share surges, classifies them, and recomputes the Leiden Ranking proportion of top-10% publications without those works under fractional and full counting. Files- affiliations_counting_rules_data_code.zip: scripts, university-level and episode-level tables, result files and figures, with a README that describes every script, its inputs and its outputs.- leiden_pairs_part1.zip to leiden_pairs_part8.zip: one array (leiden_pairs.npy, 530 MB) split by rows. Place the eight parts next to the scripts and run 00_join_leiden_pairs.py, which rebuilds the array and checks it against the SHA-256 of the original.- README.md: the README of the main archive. Universities are identified by OpenAlex and ROR identifiers. The archive contains no author identifiers and no author-level data. An episode is a statistical description of affiliation data; it is not a finding that any university paid for affiliations.