Intelligence Hours: replication data and code for the public v5 country-source correction

This data-and-code release supports the public v5 country-source correction of Intelligence Hours, dated 3 October 2026, by Kwangseob Ahn (Sejong University; ORCID 0009-0000-5299-1457). The DOI identifies a replication dataset, not an accepted journal article.

The historical panel retains 37 economies and 111 country-year rows for 2023–2025. Dated primary schedules reconstruct 105 wage references; six Canada/Japan rows retain documented employment-weighted regional provider benchmarks. The source-audit union covers 42 economies across the panel, January 2026 extension and price pilot. Forty-seven hourly references change under the disclosed effective-period, pay-unit and calendar-averaging policy. Frozen original inputs remain preserved.

Corrections distinguish Bulgaria currency alignment, ordinary basic pay and mandatory extra payments, direct hourly tariffs, dated working-time changes, regional coverage and contract categories. Japan’s January 2026 in-force range is JPY 951–1,226; Poland ordinary monthly employment is separate from civil contracts. China/India national wage ranges are withheld rather than formed from incompatible regional instruments; Nigeria’s monthly-to-hourly case remains conditional. No observed matched-median earnings series is certified: the 32-economy Kaitz product is explicitly a same-year source-ratio scenario.

Consumer-tax sensitivities use dated standard-rate or named component assumptions, not verified AI invoices. Thirteen selected 2025 employee-income profiles remain distinct from 24 OECD TaxBEN 2024 net-family-income cases. India Go is provider-stated GST-inclusive; Plus India and Indonesian inclusion remain unverified. Gross pay, employee income, employer cost, refundable credits and household benefits are not interchanged.

Regenerated analyses use HC3 cross-sectional covariance, country-clustered CR1 pooled covariance, matched-sample market/PPP comparisons, distinct-currency diagnostics, tied rollout tests and a bounded Anthropic Economic Index criterion. Independent numerical checks cover 717 assertions; portable reproduction verifies regenerated analytical files from the bundled inputs. These checks do not certify worker exposure, every legal scope, causal adoption effects or actual checkout prices.

The release includes the concise main article, separate Supplements A–F, editable DOCX and clean PDF files, figures, frozen provider responses, native-language source captures, scripts, source registers and checksum manifests. The main document contains 8550 words including tables/references; the abstract contains 81 words. The fixed pre-tax US$20 historical benchmark is separate from the website’s operational basket-v1 series. AI-assisted critique is disclosed and is not journal peer review.

Prior immutable data/code records remain at https://doi.org/10.5281/zenodo.23030431 (1.0.0) and https://doi.org/10.5281/zenodo.23101798 (1.1.0).