[vahid-ahmadi]

Corrects the default-dataset switch and rollback planning page (docs/book/assumptions/mp30k-uk-switch-plan.md), written for #1692.

The page itself is already on main: it shipped in #2386 (merged 21 October 2026), and its original .added changelog entry was released in 2.143.0. This branch has merged main, dropped the released fragment and now carries only corrections to main's copy of the page.

What this PR corrects (from its changelog fragment):

Corrected the default-dataset switch plan (docs/book/assumptions/mp30k-uk-switch-plan.md): the tests now pin Enhanced FRS 2024-25 from policyengine-uk-data 1.58.1, policyengine.py 6.10.0 already switched its UK default to Microcosm UK, an inline situation takes precedence over any dataset, unknown enum names raise an error, and the VAT survey-coverage parameter is calibrated to Microcosm UK and must be re-derived on a switch or rollback.

The claim-by-claim review behind these corrections, with sources, is in the "Ready for hub review" comment on this PR.

• Changelog fragment: changelog.d/1692-dataset-switch-page-corrections.changed.md (changed).
• Docs and changelog only; no code changes.

🤖 Generated with Claude Code

[MaxGhenis]

UK hub (Claude build session, for Max)
Ready for hub review

Head: 20cad2fc6d34b1b82c7b352b7294bc0414886a46

Context. This page already reached main, byte for byte, through #2386 (merged 21 October 2026; batch commit d50ee85), and its 1692.added.md entry is already in CHANGELOG.md under 2.143.0. I merged origin/main into this branch (no rebase, no force-push), deleted the released .added fragment, and corrected the page. The PR's net diff against main is now just those corrections plus changelog.d/1692-dataset-switch-page-corrections.changed.md. Suggested title: "Correct the default-dataset switch plan".

Is the plan still worth having? Partly overtaken: policyengine.py 6.10.0 (21 October 2026) already switched its UK default to Microcosm UK. The page now says so, and says it doesn't record whether these gates were run for that switch. The plan still applies to switching this repository's test pin, which stays on the Enhanced FRS. I recommend keeping it.

Review. I had a verifier extract and check every claim, checked each correction myself, and then ran independent adversarial reviews. Rounds 1 and 2 asked for changes, and round 3 approved; its one optional precision point is folded into this head.

Corrected (13)

1. Resolution order: an inline situation is used before dataset= and the environment variable; otherwise get_default_dataset_url raises a ValueError (policyengine_uk/simulation.py:181-189,349-363).
2. The test pin is enhanced_frs_2024_25.h5 from policyengine-uk-data 1.58.1, pinned by commit and SHA-256 in policyengine_uk/tests/certified_dataset.yaml, not the 2023-24 1.40.3 file.
3. The tests set the environment variable only when a token is present and the variable isn't already set (policyengine_uk/tests/conftest.py:118-121).
4. policyengine.py 6.10.0 switched its UK default to microcosm_uk_2024_25 (its CHANGELOG and manifest). It did so under its own certification, which records an accepted UC caseload-floor exception.
5. Base years now match: both are 2024-25.
6. String enum columns are matched by member name, and an unknown name raises a ValueError (policyengine-core enums/enum.py:87-107). Integer columns are not checked (simulation.py:215-219).
7. variables/input/ holds many inputs, but the loader sets any column named after a variable (build_from_multi_year_dataset, simulation.py:1189-1197). Examples defined elsewhere: tenure_type, household_weight, would_claim_uc, *_reported.
8. Criterion 4: test_latest_data_smoke.py:29-31 loads a hard-coded, unpinned enhanced_frs_2023_24.h5, and the other two named tests use inline situations or a hand-built dataset. The criterion now names the pin and the reform fixtures (tests/microsimulation/reforms_config.yaml).
9. Both hf:// loading (core tools/hugging_face.py:139) and the pin's materialise (utils/aggregate_impact/dataset.py:155-159) request model repositories, and load_pin accepts no repository-type key. Microcosm sits in a dataset repo, so pinning it needs a loader change.
10. Rollback is not code-free: gov.simulation.vat.survey_consumption_coverage is calibrated to one Microcosm cut (20261002T230158Z-5c6b3f68). On the Enhanced FRS it puts VAT 38% (uk-data 1.56.16) or 44% (1.58.0) above receipts (parameters/gov/simulation/vat/survey_consumption_coverage.yaml).
11. Per-environment defaults: the env var for direct use, the committed pin for tests, and the bundle manifest for policyengine.py.
12. Rollback steps and the Acceptance text now include re-pinning and re-deriving dataset-calibrated parameters.
13. Replaced the released .added fragment with a .changed fragment.

Checked and correct (unchanged)

• DEFAULT_DATASET_ENV_VAR is set at tax_benefit_system.py:103.
• The Microsimulation() deprecation is in README.md:6.
• UKSingleYearDataset and UKMultiYearDataset are in data/dataset_schema.py, and _pre_encode_enum_columns is at simulation.py:191-222.
• Microcosm publishes microcosm_uk_2024_25.h5 and calibrates and certifies its own releases.
• microplex-uk is archived, and its discussion Add income-based JSA #2 exists.
• Define PolicyEngine-UK switch and rollback path for mp-30k-uk #1692 and Align UK model/data pipeline with US patterns: reported-anchoring, second-stage imputation, and downstream bugs #1621 are open, and all links return 200.

Uncertain

• That hf:// cannot load a dataset-repo file is deduced from the code. I didn't run it.