Replication package: a cross-funder event table of targeted heritage-digitalisation funding programmes

This dataset supports replication of the construction of the event table used in a preregistered study of whether targeted grant programmes redirect the direction of knowledge production. It contains the frozen rule implementation, the resulting event table, the complete manual-verification apparatus, and the preregistration documents. Contents: preregistration and pre-estimation addendum; six pipeline scripts (candidate identification, consolidation, assembly, verification-sample generation, inter-rater statistics); the event table (4,079 candidate event groups; 2,954 after excluding NIH container-weak groups), the rejected-candidate log (16,610 groups), the synonym-table seed, the 119-group stratified verification sample with round-1 labels from both raters, the verification guide, the round-1 adjudication template, and the round-2 re-labelling templates. Verification status at deposit: round 1 of dual-rater verification is complete (Cohen's kappa = 0.465 < 0.70); the preregistered re-labelling round with a written clarification memo is in progress. The record deliberately contains no outcome data, no effect estimates, no manuscript text, and no figures: at the time of deposit the preregistration lock conditions had not been fully met, and the lock prohibits any effect estimation before they are. Derived outcomes and estimates will be added in a later version of this record. All raw funder panels are public archives (CORDIS, NSF, NIH RePORTER, UKRI Gateway to Research) and are not redistributed; download instructions and pipeline order are given in the included README. The record is deposited anonymously, without author names, institutions, or journal information, in keeping with double-anonymized peer review. Version 2.0 (2026-10-29). Adds the derived outcome panels (analysis sample, cross-funder comparison panel, full-corpus outcome coding, purified corpus metadata), the publication-to-funder acknowledgement crosswalk (provided as four row-wise zstd-compressed parquet shards, reassemble by concatenation), the frozen verified event table, all effect estimates (results/results_est.json), the post-review replication outputs including cluster-bootstrap equivalence intervals and measurement-sensitivity analyses (results/revision3_calc.json), the measurement gold standard with texts, and the complete estimation code (outcome coding, treatment assignment, estimation, and post-review replication scripts). The reproduction pipeline order is given in the included README.