Validated use: per-package counts

How often each R, Python and Stata package is loaded by the code deposited with published papers, at journals whose data-and-code policy the Social Science Data Editors record as actively verified.

4,565 packages · 13,245 deposits with analyzable code · 13,985 deposits collected · built 2026-09-18

A count here is the number of deposits whose code loads the package. Adding one to it takes a paper published at a journal that checks its authors’ code, which is what makes these counts harder to inflate than download counts. The project page makes that case; this file documents what is in the tables and how to read them.

Scope

The deposits come from two repositories that hold different disciplines. Zenodo’s verified collections are economics, and Harvard Dataverse’s journal collections are mostly political science.

repository	deposits	with analyzable code
Zenodo (economics)	1,543	1,358
Harvard Dataverse (political science)	12,442	11,887
total	13,985	13,245

Counts pool the two. usage_by_package.csv also carries the split, in n_deposits_zenodo and n_deposits_dataverse, because the two are very different sizes and a pooled figure alone would hide that.

Both repositories were collected with the same rules in 2026: every code, notebook and knitr file and every dependency manifest, with the deposit’s own directories, and code recovered from archives too large to download. An archive that could not be read is counted in summary.json rather than dropped. The two collections still differ in discipline, which the per-source columns keep visible.

Files

file	rows	contents
usage_by_package.csv	4,565	per-package deposit and call counts, pooled and split
usage_by_package_year.csv	14,670	the same by deposit year
usage_by_collection.csv	22,838	the same per journal or community
usage_by_function.csv	20,942	package → function, where the source names one
unknown_names.csv	2,787	names called in code that resolve to no registry
remote_installs.csv	473	what deposits install from outside their registry, and from where
downloads_vs_use.csv	43,918	each package’s validated use beside its registry’s download count
language_presence.csv	24	deposits containing each language, per repository
mentions.parquet	14,853,082	every mention: package, function, file, line, snippet
files.parquet	447,289	the provenance spine every mention joins to
declared_dependencies.parquet	33,874	what manifests declare: shipped, locked or asked for
environment_signals.parquet	19,983	R, Python and Stata versions, and the OS, where a file says
environment_coverage.json		deposits stating each signal, over deposits that could
summary.json		corpus counts the tables are shares of

mentions.parquet is the record every count above is a sum of, and it is here so a reader who disagrees with a decision made upstream can recount without re-parsing 200,000 files. It is 94 MB; the CSVs are 150 KB.

declared_dependencies.parquet and environment_signals.parquet are sparse and answer a different question: not what the code loads but what version of it the deposit shipped, and what ran it. Most deposits say nothing at all, so read these next to environment_coverage.json, which gives per signal the deposits that said something over the deposits that were in a position to.

usage_by_package.csv

• package, language, ecosystem: the resolved package and its registry

• n_deposits: deposits loading it, counted once per deposit

• n_files, n_mentions: files, and raw calls

• n_deposits_at_risk: the denominator for share_of_deposits, meaning deposits that hold an analyzable file in that language, or that yielded a reference in it. The second clause is what covers literate documents, where the file is a notebook and the code inside it is Python. It differs by language, so a Stata share and an R share are not shares of the same thing

• share_of_deposits: n_deposits divided by n_deposits_at_risk

unknown_names.csv

Names that code uses and that resolve to no registry, unfiltered. Install and inquiry lines are excluded: ssc install x states a dependency and is not a call. Some rows are false positives, and some are programs a deposit defines for itself. Pruning the list by hand would put a judgement call inside a file whose value is that you can check every row of it.

• name, language

• n_deposits, n_mentions: deposits using the name, and raw uses. Rank by deposits: one deposit calling something six hundred times is one user of it

• n_deposits_defining: how many deposits define a Stata program of this name for themselves. A name many authors independently give a helper is more likely one here too, with a program define the lexer did not reach, than it is software nobody indexed

remote_installs.csv

What deposits fetch from somewhere other than their language’s registry: remotes::install_github("user/repo"), net install x, from(URL), pip install git+https://.... It is the only record a deposit leaves of where off-registry software lives.

• name, language, host: the package and the host it is fetched from

• in_registry: the registry lists the name anyway, so this is a development version of a registered package rather than software the registry lacks

• n_deposits_installing, n_deposits_loading: deposits with the install line, and those among them that go on to use the package

downloads_vs_use.csv

Validated use beside download counts, for every package that SSC, CRAN or PyPI has a count for or that the corpus uses. The download counts are not ours: SSC’s monthly hits behind ssc hot, CRAN’s from one mirror through cranlogs, and PyPI’s for its fifteen thousand most downloaded projects from hugovk/top-pypi-packages, each pinned by date and digest. The windows differ, so compare ranks within one registry, not counts across them.

• package, language, ecosystem

• n_deposits: deposits using the package; 0 for a counted package this corpus never uses

• downloads: the registry’s count; empty for a used package it has none for

• in_registry_counts: the row belongs to the registry’s own list of counted packages, as against a count fetched because the corpus uses the package

• reverse_dependencies, in_task_view: CRAN only. How many packages install this one as a dependency, directly or through others, and whether CRAN’s Econometrics or Causal Inference task view lists it

Licence

CC0.

Regenerating

uv run python scripts/build_tally.py     # needs the collected corpus
uv run python scripts/release_tally.py

Produced by softverse.