Skip to content

Add DWP's young people and work review facts: PIP and UC health element by age, and the final report's Tables 2 and 3 (chronicle#442) - #443

Merged
juaristi22 merged 2 commits into
mainfrom
uk-young-people-benefits-442
Nov 18, 2026
Merged

juaristi22 merged 2 commits into
mainfrom
uk-young-people-benefits-442

Conversation

@juaristi22

@juaristi22 juaristi22 commented Nov 18, 2026 •

Copy link
Copy Markdown
Collaborator

Closes #442. Two packages add 117 facts from DWP's publications for the Young People and Work review, under the Open Government Licence v3.0, written by a generator script from pinned bytes; no microdata is read or stored. Items 3 and 4 need no new package (see below).

The interim report's Table 28 (#442 item 1)

dwp-young-people-pip-lcwra-projections-2026 (88 facts), from the analytical annex data tables of Young people and work: interim report (published 28 May 2026). The pinned file is the edition GOV.UK replaced on 10 August 2026, when "an error in the Excel formula" was corrected in Table 23 only.

Table 28, "PIP and UCH-LCWRA caseload and expenditure projections, aged 16-24", prints every financial year from 2019-20 to 2040-41, not only the three years the issue lists. All four rows are ported:

  • UC health element (LCWRA) caseload and expenditure;
  • PIP caseload and expenditure.

Semantics:

  • Concepts: DWP's own dwp.benefit_caseload_count and dwp.benefit_expenditure_amount, as dwp-benefit-expenditure-caseload-spring-2026 uses them, with filter health_and_disability_benefit.
  • Who: age_band 16_to_24, England and Wales ("Population: England and Wales").
  • People, not households: the annex defines UC Health as "claimants in receipt of the 'limited capability to work' element", so the caseload counts claimants (person / claimant). Microcosm's dwp.uc.households_lcwra_element counts households.
  • Outturn and projection: the table's period line reads "Period: 2026-27 to 2040-41", the annex's "fifteen-year caseload projections". The annex's forecasts "are based on data from Stat-Xplore and Autumn Budget 2025 benefit caseload and expenditure tables", which still forecast 2025-26.
    • 2025-26 to 2040-41 are DWP's projections (model_output, source_projection), 2025-26 typed as dwp-benefit-expenditure-caseload-spring-2026 types it.
    • 2019-20 to 2024-25, outturn years in those tables, are the history (administrative, observation).
  • Rounding: caseloads to the nearest 1,000; spending to the nearest £10m (UC health) and £100m (PIP), scaled to pounds. Neither the table nor the annex names a price basis, so none is asserted.
  • Caveat carried: the projections "have used a simple methodology and should be treated as an indication only", and "may not align with published benefit expenditure and caseload tables".

Checks: the annex's findings quote the table, and the generator asserts each figure:

  • PIP spending of £1.3bn in 2019-20 and £3.3bn in 2024-25;
  • £6.5bn and 700,000 young people in 2031-32;
  • 877,000 and £9.2bn in 2040-41;
  • the LCWRA caseload of 229,000 and £750m in 2040-41.

The workbook's cover prints "Published:28 May 2025", a slip for 2026.

The final report's data tables (#442 item 2)

dwp-welfare-state-to-working-state-tables-2026 (29 facts), from the data tables of From a Welfare State to a Working State: a new settlement for the next generation (17 November 2026).

  • No caseload series: the workbook has three tables, and none carries young people's PIP or UC health caseloads or spending. The annex (section 9) takes its 2036/37 PIP and DLA caseloads from the interim report's forecasts, which Table 28 is.
  • What it ports: the two comparators UK facts for young people's health-related benefits by age, and HMRC's employer NICs reliefs for young employees #442 asks for. Both are the review's "internal modelling" (model_output, source_projection).
  • Caveat carried: the guidance says the analysis "should not be regarded as official government analysis or as representing agreed government policy".

Table 2, annual cost and saving per policy area in 2036/37, £m, 2026/27 prices (7 facts, price_basis 2026_27_prices). The annex says "The geographical coverage of the costs vary", so each area carries the coverage the annex states:

  • England: health and education.
  • Great Britain: the Opportunity and Support Guarantees.
  • The United Kingdom: welfare savings, "PIP and UC expenditure across the UK".
  • Approximate stamps: three rows carry concept relation approximate.
    • Youth hiring incentives: Great Britain grants alongside England-only apprenticeships, stamped Great Britain.
    • Fiscal offsets: no coverage stated, stamped the United Kingdom.
    • The net total: the sum of every area, stamped the United Kingdom.
  • "[low]": "Local system reform" prints "[low]", "a figure that rounds to zero but is not zero". It emits no fact, and the other rows sum to the total of -£200m.

Table 3, the NEET rate projection, UK, 2026/27 to 2036/37 (22 facts; the title's "2036/27" is a slip). It holds the "Projected baseline NEET rate" and the "Projected NEET rate with report recommendations" for 16 to 24 year olds, filter neet_rate_projection.

  • Baseline: it starts from "a central, seasonally-adjusted NEET projection scenario from Youth Futures Foundation (YFF) SARIMAX modelling", adjusted by DWP and DfE.
  • Checks: the two agree in 2026/27, and the recommendations lower the rate every later year, to the annex's "around 6.8%" in 2036.

Table 1 (past schemes' average annual investment) is not ported: the table gives no periods.

Items 3 and 4

  • Item 3, LCWRA households by claimant age: Stat-Xplore's Households on Universal Credit cube has no claimant age field. Its age fields are the number of children by age and the age of the youngest child. People on Universal Credit has age but no capability for work field, and no Universal Credit cube crosses the two. The nearest fact, people on UC by age and conditionality regime, is already in Chronicle.
  • Item 4, HMRC's employer NICs reliefs: already in Chronicle as hmrc-employer-nics-age-reliefs-january-2026 (UK facts for four November events: young people and work, the PIP review, the January 2027 price cap and the Scottish Budget 2027-28 #420 item 9).
    • It carries Table 2's cost of NIC-NS1 and NIC-NS2, 2020-21 to 2025-26, with 2025-26 as HMRC's forecast, and Table 5's employers claiming each.
    • It quotes HMRC's note that the apprentice relief takes precedence for apprentices under 21.
    • January 2026 is still HMRC's latest edition, and the pinned workbook is the file GOV.UK serves.

Checks

  • Generator: --check reports the committed YAML up to date.
  • Tests:
    • test_source_uk_young_people_work_review.py (12) re-reads both workbooks with an ODS reader of its own and matches every fact to its cell;
    • test_dimension_value_register.py and test_issue_376_one_label_per_value.py (55);
    • test_chronicle_bundle.py with the whole-bundle test deselected (34);
    • test_chronicle_boundaries.py, test_source_package_alias_drift.py, test_chronicle_source_table_year.py and test_chronicle_artifacts.py (215);
    • test_chronicle_source_package.py (144);
    • test_issue_412_uk_area_inputs.py with the boundaries test (92).
  • Register: the static check finds no problems.
  • Bundle pins move by the packages' own counts:
    • facts 2,312,590 (+117), packages 517, tables 459;
    • periods 619: Table 28's 2032-33 to 2040-41 are new financial years;
    • dwp 274,652, people +66, government +51;
    • England and Wales +88, the United Kingdom +25, Great Britain +2, England +2.
  • Unchanged: semantic duplicate keys (4,135) and warnings (92). The period, entity and source pins each sum to the fact count. The whole-bundle test runs in CI.
  • R2: both raw workbooks are uploaded with chronicle publish-raw and read back with chronicle verify-raw.

🤖 Generated with Claude Code

- dwp-young-people-pip-lcwra-projections-2026 (88 facts): the interim
  report's analytical annex Table 28, the PIP and UC health element (LCWRA)
  caseloads and spending of 16 to 24 year olds in England and Wales, every
  year 2019-20 to 2040-41. Years from 2026-27 are DWP's projections, as the
  table's period line says. The facts sit on DWP's own
  dwp.benefit_caseload_count and dwp.benefit_expenditure_amount.
- dwp-welfare-state-to-working-state-tables-2026 (29 facts): the final
  report's data tables 2 (the recommended policies' annual cost or saving by
  policy area in 2036-37, 2026-27 prices, each area at the coverage the
  annex states) and 3 (the NEET rate projected with and without the
  recommendations, 2026-27 to 2036-37). Both are the review's modelling.

Both raw workbooks are in R2 and read back. Bundle pins: facts 2,312,590
(+117), packages 517, tables 459, periods 619 (fiscal years 2032-33 to
2040-41 are new), dwp 274,652, people +66, government +51.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

@vahid-ahmadi vahid-ahmadi left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for this. The packages are carefully built, and the values hold up against the workbooks.

Checked against the committed .ods files with a separate reader:

  • All 117 facts match their source cells: Table 28 88/88, Table 2 7/7, Table 3 22/22. The guard cells and headers match too.
  • value_scale 1e6 on £m. Caseloads are persons. Table 2 savings are negative.
  • The Table 2 rows sum to the -£200m total with "[low]" omitted.
  • NEET rates are stored as percent, consistent with ons.lfs_neet_rate.
  • The manifest sha256 values match the bytes. The Table 28 URL's Last-Modified (10 Aug 2026) and size match the pinned corrected edition.
  • --check passes, and test_source_uk_young_people_work_review.py plus the register and alias tests pass locally.
  • The bundle pin deltas add up (117; UK +25, E&W +88, GB +2, England +2; person +66, government +51; 2032-33 to 2040-41 new).

One question on the 2025-26 boundary in Table 28.

  • How it is tagged: scripts/build_uk_young_people_work_review_packages.py:486 tags 2025-26 (column I, table_28.caseload.fy2025 / table_28.expenditure.fy2025) as administrative/observation because the table's period line is "Period: 2026-27 to 2040-41".
  • What the workbook says: that line marks where DWP's projection starts. The table does not label 2025-26 as outturn; A5 says only "Source: DWP administrative data".
  • What the annex says: the annex sentence the package header quotes (packages/dwp/young_people_pip_lcwra_projections_2026/source_package.yaml:27) reads "PIP and UC-H / LCWRA caseload and expenditure forecasts … are based on data from Stat-Xplore and Autumn Budget 2025 benefit caseload and expenditure tables". In those tables 2025-26 is a forecast year.
  • Chronicle's own tagging: dwp-benefit-expenditure-caseload-spring-2026 tags 2025 as model_output.
  • Why it matters: if 2025-26 is partly forecast (spending especially, with the report out in May 2026), consumers reading observation as outturn would treat it as published actuals.
  • Suggestion: consider tagging 2025-26 as model_output/source_projection, or noting the uncertainty on those two record sets.
  • Wording: docs/data-sources/uk-young-people-work-review-442.md:60-61 and the package header say the annex "bases the history" on those tables, but the annex sentence describes the forecasts.

CI ("Chronicle bundle", "Chronicle surface") was still pending at the time of review.

Review drafted with AI assistance and checked against the code at 2081100.

Review of 2081100 (#443): the annex says its
"forecasts ... are based on data from Stat-Xplore and Autumn Budget 2025
benefit caseload and expenditure tables", and those tables still forecast
2025-26, as dwp-benefit-expenditure-caseload-spring-2026 types the year.
So 2025-26 joins the projections (model_output, source_projection), a year
before the table's period line, and 2019-20 to 2024-25 stay the history.
The notes now quote the annex sentence as the forecasts' source, not the
history's. Fact values and bundle counts are unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@juaristi22

Copy link
Copy Markdown
Collaborator Author

Thanks for checking every fact against the workbooks.

On the 2025-26 boundary: agreed, and fixed in 0cd9942. The annex's forecasts rest on the Autumn Budget 2025 tables, which still forecast 2025-26, so 2025-26 now joins the projections (model_output / source_projection), as dwp-benefit-expenditure-caseload-spring-2026 types the year. 2019-20 to 2024-25 stay the history. The notes, package header and docs now quote the annex sentence as the source of the forecasts, not of the history, and the PR description says the same. Fact values and bundle counts are unchanged.

🤖 Generated with Claude Code

@juaristi22
juaristi22 marked this pull request as ready for review November 18, 2026 12:39
@juaristi22
juaristi22 merged commit 15040ea into main Nov 18, 2026
3 checks passed
juaristi22 added a commit to PolicyEngine/microcosm that referenced this pull request Nov 18, 2026
…le 15040ea5

Part 3 of the UC health fix on #1663, on
PolicyEngine/chronicle#443, which María merged on 18 November 2026 ("post
it and continue with the steps that follow from 443").

- dwp.uc.people_aged_16_to_24_lcwra_element binds DWP's Young people and
  work: interim report, Table 28: people aged 16 to 24 in England and Wales
  with the UC health element. It takes the calendar-2025 window of 2024-25
  (outturn) and 2025-26 (DWP's projection), María's rule of 2026-09-22 for
  fiscal-year facts: 176,000. The model counts claimants and partners aged
  16 to 24 holding uc_limited_capability_for_WRA in units paid UC with the
  LCWRA element, in England and Wales.
- uc_take_up passes the same count to the young LCWRA fit, from the
  vendored dwp_uc_young_lcwra_people_facts.json. The ratio is solved on
  England and Wales's units and applies across Great Britain.
- tools/generate_uk_target_references.py lets a UC row declare
  calendar_year_window in place of the monthly calendar-year average.
- The re-pin to 15040ea5 has 2,160,350 facts. Every 4ce631f5 fact line is
  byte-identical, and the 117 new lines are #443's. The vendored
  resources, the DESNZ constituency resource, the national and local
  references and membership, the local census, the validation levels, both
  parity receipts, the coverage manifest and the H2 fixture are regenerated
  in the runbook's order. The national surfaces gain the one row and
  nothing else moves; the local surfaces change only their feed identity
  and derived fact keys.
- Pins: 657 targets, 595 registry rows, 412 unmapped declarations and 1,714
  active references; the feed pin tests restate 15040ea5.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

UK facts for young people's health-related benefits by age, and HMRC's employer NICs reliefs for young employees

2 participants