# A07 · Who Receives the Automation Dividend?

Dr. Freedom Cheteni · Journal for School Superintendents · The Research University

Version 1.0 · prepared 15 October 2026

This research package accompanies a new prospective primary-study protocol corresponding to conceptual article A02. No review or acceptance is inherited. No schools or teachers have been recruited and no intervention effects are reported. The 48-school scenario is provisional; section 14a and the analysis plan define a mandatory independently reviewed power/quality activation gate before main enrollment.

Completed analyses are (1) an integrity and descriptive feasibility audit of official BLS ATUS 2025 respondent/activity files, and (2) 10,000-replicate artificial matched-pair design calculations in nine scenarios. Original BLS files are downloaded directly from the agency, not redistributed here. The September 2026 User's Guide, June 2026 dictionary and 2018 occupation mapping are pinned by hash in source-audit.json. The 2025 diary shutdown gap is explicit. Source audits are not human peer review.

## Reproduce

Use Python3.12 and requirements.txt for the simulations. The audit uses only the standard library. Install dependencies in an isolated environment. Download the two official files into a private working directory:

- https://www.bls.gov/tus/datafiles/atusresp-2025.zip
- https://www.bls.gov/tus/datafiles/atusact-2025.zip

```
python research_tools.py audit --respondent /absolute/path/atusresp-2025.zip --activity /absolute/path/atusact-2025.zip --output reproduced-atus-audit.json
python research_tools.py simulate --seed 20261015 --replicates 10000 --output reproduced-design-analysis.json
python test_research_tools.py --source-dir /absolute/path/source-directory
```

For the last command, place copies named respondent.zip and activity.zip in the source directory, or use --respondent and --activity individually. Without source paths the unit/content tests run and reproduction is reported as skipped, not passed. The audit fails closed if either ZIP hash differs from the inspected source version. Obtain and explicitly audit a new version rather than bypassing the pin to claim an identical reproduction.

The tool never transmits input records. It does not implement the future primary analysis, imputation, actual allocation, quality assessment or final power gate. Those implementations and their independent review are prerequisites for study activation. No primary participant records should be published with this package.

## Files

- article.md: complete original protocol with 17 references.
- study-specification.json: design, actual-data status and activation requirements.
- instruments.md: voluntary diary, timetable charter, blinded synthetic quality tasks and safety process.
- analysis-specification.md: estimands, missingness, quality guardrail, activation gate and exact executed sensitivity assumptions.
- source-audit.json: URLs, versions, source hashes, inspected variables, citation verification and access limits.
- atus-feasibility-audit.json: completed secondary-source results.
- design-analysis.json: completed artificial-data results with Monte Carlo uncertainty.
- research_tools.py and requirements.txt: restricted reproducible audit and simulation.
- test_research_tools.py: unit, source-accounting, content and optional real-file reproduction checks.

Citations preserve source authorship. No open-access or Creative Commons republication license is asserted for the cited Science/QJE papers. Reproduction data remain with BLS; original source PDFs are not repackaged. The journal's actual author approval and licensing records govern licensing of the new article and research materials.