Validated use: per-package counts of software loaded by social science replication code

How often each R, Python and Stata package is loaded by the code deposited with published papers, at economics and political science journals whose data-and-code policy an editor verifies before publication. 3,223 packages · 8,349 deposits with analyzable code · 69 journal collections. Why count this way. A metric used to allocate credit is a signal, and a signal is informative in proportion to what it costs to produce. A download is one fetch of a file, which a build server installing on every commit supplies in quantity and a faster release schedule supplies more of. A mention in an article's prose is one sentence. Adding one to a count here takes a paper accepted at a journal that checks its authors' code, with the package loaded in that code. Scope. Two repositories holding two disciplines: 1,454 deposits from Zenodo's verified economics collections and 7,231 from Harvard Dataverse's political science journals. Counts pool both and usage_by_package.csv carries the split beside every pooled total, since the two are very different sizes. What the counts do not show. These counts say a package was loaded by code in the deposit. They do not say the code ran. Authors often leave older scripts in a deposit, and a script can load a package inside a branch that never executes. Whatever an author kept out of the deposit cannot be counted at all. The Harvard Dataverse material also came from a January 2024 scrape that kept only .do, .r and .py files, so a package used mainly inside a notebook is under-counted on that side; the release checks this by recomputing the whole ranking on those three extensions and fails if the order moves. Files. usage_by_package.csv is the main table. usage_by_package_year.csv and usage_by_collection.csv give the same counts by deposit year and by journal. unknown_names.csv lists names used in code that resolve to no registry, unfiltered. summary.json holds the corpus counts the shares are taken against, and README.md is the data descriptor with column definitions. Stata resolution uses the command-to-package index deposited separately at 10.5281/zenodo.21926100. The most loaded package in this corpus is estout, in 2,440 deposits.