Disregard the whole income and capital of Housing Benefit claimants on IS, income-based JSA or income-related ESA

[MaxGhenis]

Fixes #1993.

Merge order. This PR lands after #1896, as the PE-UK merge train set. #1896 merged on 2026-10-02 (3525263). The head (2febacb) is the reviewed b6d45a5 plus #2048's test fixes and merge commits of main (latest 2c94286) and of #1896's head. Against main it changes only the files listed under "Change" and #2048's.

Problem

housing_benefit_applicable_income added income_support, jsa_income and esa_income to a Housing Benefit claimant's income and tapered it, and housing_benefit_assessable_capital counted their capital. In law, a claimant on one of these benefits has the whole of their income and capital disregarded, so an eligible claimant gets maximum Housing Benefit before the benefit cap: the eligible rent, capped at the LHA rate where it applies, less non-dependant deductions, floored at zero (reg 70).

The hand-computed YAML cases below are real runs of this branch and main 8b2e6e1 with #1896 merged in. They gave the same values on main 8b2e6e1 and 7b9fc37 alone. Each is a council tenancy with no non-dependants, so maximum Housing Benefit is the rent.

Case	main: capital, applicable income → HB	this PR
A. Lone parent aged 30, child aged 3, Income Support £5,000, earnings £3,000, savings £12,000, rent £6,240 (2025)	£12,000 (tariff £1,248), £8,836.62 → £3,607.15	£0, £0 → £6,240
B. Single aged 40, income-based JSA £3,500, earnings £1,000, savings £12,000, rent £5,200 (2025)	£12,000, £5,394.81 → £4,804.33	£0, £0 → £5,200
C. Single aged 40, income-based JSA £3,500, no earnings, savings £20,000, rent £5,200 (2025)	£20,000: over the £16,000 limit, so £0	£0 → £5,200
D. Couple aged 32 and 30, the partner on income-based JSA £4,000, claimant earns £12,000, rent £6,240 (2025)	£0, £15,293.61 → £1,189.00	£0, £0 → £6,240
E. Mixed-age couple aged 70 and 60, the younger on income-related ESA £6,000; State Pension £12,000, private pension £15,000, savings £10,000, rent £6,240 (2026)	£10,000, £29,635.59 → £0	£0, £0 → £6,240
F. As B, with contribution-based JSA (control)	£12,000, £5,488 → £4,232.80	unchanged
G. As E, with contributory ESA only (control)	£10,000, £29,635.59 → £0	unchanged

E sets Guarantee Credit to nil, so the result does not come from the Pension Credit passport. F is fully hand-computed with the disregard (£260) and applicable amount (£4,000) given: 5,200 − 0.65 × (3,500 + 1,000 + 1,248 − 260 − 4,000) = £4,232.80.

Law

The texts were read from legislation.gov.uk CLML (/data.xml) on 2026-10-01: the current revised text, and the point-in-time text at 2026-04-06, which is identical for these paragraphs. Copies are in the research folder.

SI 2006/213 (Housing Benefit Regulations 2006).

• Sch 5 para 4 (sums disregarded in the calculation of income other than earnings): "Where a claimant is on universal credit, income support, an income-based jobseeker's allowance or an income-related employment and support allowance the whole of his income."
• Sch 5 para 5: "Where the claimant is a member of a joint-claim couple for the purposes of the Jobseekers Act and his partner is on an income-based jobseeker's allowance, the whole of the claimant's income."
• Sch 6 para 5 (capital to be disregarded): the same four benefits, "the whole of his capital". Sch 6 para 6 is the joint-claim counterpart.
• The income-related ESA words date from 27.10.2008 (SI 2008/1082 regs 25(a) and 26(a)) and "universal credit," from 28.10.2013 (SI 2013/2070 reg 7(8) and (9)).
• Reg 2 says what "on" means:
  • "person on income support" means "a person in receipt of income support";
  • reg 2(3) and (3A): a person is on income-based JSA or income-related ESA "on any day in respect of which [it] is payable to him", and also on some days when it is not paid (a sanction or disqualification, a waiting day, a loss-of-benefit restriction);
  • reg 2(3B): a person is on universal credit on any day they are entitled to it, "whether it is in payment or not".
• Reg 5(1)(b): these regulations apply to a person who "has attained the qualifying age for state pension credit if he, or if he has a partner, his partner, is a person on universal credit, on income support, on an income-based jobseeker's allowance or on an income-related employment and support allowance". SI 2006/214 reg 5(2) disapplies the pension-age regulations in the same case. So a mixed-age couple whose younger member is on income-related ESA is assessed under SI 2006/213 and passported. DWP circular A9/2019 says such a couple "can continue to receive working age HB".
• Reg 82(1): "In the case of a couple ... a claim shall be made by whichever one of them they agree should so claim".

Northern Ireland: SR 2006/405 has equivalent passport provisions. The paragraphs below use the same words, with "Jobseekers Order" for "Jobseekers Act". Reg 2(3) and (3A) cite the NI sanction and fraud legislation in place of the GB Acts:

• Sch 6 para 4 and para 5 (income);
• Sch 7 para 5 and para 6 (capital);
• reg 2(3) to (3B) and reg 5(1)(b); SR 2006/406 reg 5(2).

The NI ESA words date from 27.10.2008 (SR 2008/378) and "universal credit," from 8.5.2018 (SR 2018/92). NI's Sch 6 para 6, a 2003 rate-rebate saving, has no GB counterpart and is not modelled.

Change

• in_receipt_of_income_support_jsa_ib_or_esa_ir (new, BenUnit, bool): true when income_support, jsa_income or esa_income is positive. It follows in_receipt_of_guarantee_credit: receipt means a positive amount paid.
• housing_benefit_applicable_income: nil when that holds (Sch 5 para 4). Adds documentation and reference.
• housing_benefit_assessable_capital: nil when that holds (Sch 6 para 5). Tariff income is then nil and the capital limit is met, so housing_benefit_tariff_income and housing_benefit_eligible need no change.
• Docs: one sentence in the "How PolicyEngine computes Housing Benefit" cell of housing-benefit.ipynb.
• Existing tests: three cases in housing_benefit_applicable_income.yaml used Income Support £2,250 as ordinary income. That income now passports, so the cases are rebuilt on £3,250 of earnings, one also with £2,000 of working tax credit, and no Child Benefit (which Sch 5 para 65 disregards). Their expected values (£2,500, £4,500 and £0) are unchanged, and the tests still check what they were written for (pension contributions halved from earnings and capped, other income added).

Modelling choices

• Receipt, not "on". An amount paid above zero stands in for "on". The model cannot see the reg 2(3) and (3A) days when nothing is paid.
• The benefit unit. The model holds these awards for the benefit unit and cannot tell which member claims Housing Benefit, so it applies the passport whichever member is on the benefit. A couple choose who claims (reg 82(1)), and para 5 covers a joint-claim couple.
  • Replace generic child and adult flags with each programme's legal definitions #1896 has merged, but the receipt test does not need its claimant-and-partner definition in practice. jsa_income and esa_income add up the members' reported awards, so a dependant's own award would count. But in the Enhanced FRS every member who reports Income Support, income-based JSA or income-related ESA is a claimant or partner (is_uc_claimant), in 2025 and 2026. Restricting the sums to is_claimant_or_partner is a follow-up (Count only the claimant's and partner's legacy awards in means tests and passports #2027 scopes these award readers).
• No age condition. This follows reg 5(1)(b). A handful of wholly pension-age benefit units in the Enhanced FRS report income-related ESA or income-based JSA. None of them receives Housing Benefit, so the choice changes nothing there. If any did, the passport would follow the model's award; the age conditions belong in the ESA and JSA variables.
• The universal credit limb is left out, as in Take Housing Benefit earnings disregards from net earnings only #1908's para 12 branch:
  • the model never pays Housing Benefit and Universal Credit to the same family (housing_benefit_eligible; 0 such benefit units on the Enhanced FRS in 2025, 2026 and 2027);
  • universal_credit depends on Housing Benefit through benefit_cap_reduction, so reading it here would be circular.
• From 2026-27 only income-related ESA passports. gov.dwp.income_support.active and gov.dwp.JSA.income.active are false from 1 April 2026, so the model pays no Income Support or income-based JSA from 2026-27.

Interplay with open PRs

All figures are real runs of scratch merges on the same Enhanced FRS copy, for 2026 unless stated. They were run on the earlier base (main 7b9fc37, branch b6d45a5), before #1926 changed the LHA order, and have not been re-run on today's main.

• Take Housing Benefit earnings disregards from net earnings only #1908 (earnings disregards; Sch 4 para 12). The two PRs merge cleanly in either order.

  • There is no double counting. With both merged, Take Housing Benefit earnings disregards from net earnings only #1908 changes no passported family's Housing Benefit (0 records differ from this branch alone): applicable income is nil whatever the disregard.
  • Take Housing Benefit earnings disregards from net earnings only #1908 alone lowers working-age HB by £3.66m for 13.3k units, and 13.26k of them (−£3.64m) are on these benefits. With both PRs merged, those families are back at maximum: on Take Housing Benefit earnings disregards from net earnings only #1908's branch this PR adds +£20.49m for 13.3k units, against +£16.85m on main.
  • Suggested follow-up for whichever lands second: Take Housing Benefit earnings disregards from net earnings only #1908's para 12 branch tests add(IS, JSA, ESA) > 0 & ~pension_age. It could use in_receipt_of_income_support_jsa_ib_or_esa_ir and drop ~pension_age, since reg 5(1)(b) puts a mixed-age couple on these benefits under SI 2006/213. That changes only the reported disregard, not Housing Benefit.

• Key the Housing Benefit Guarantee Credit passport on receipt #1927 (Guarantee Credit passport on receipt). Both PRs edit the same two passport lines, so whichever lands second has a small conflict.

  • Resolution: keep Key the Housing Benefit Guarantee Credit passport on receipt #1927's passported = any_over_SP_age & benunit("in_receipt_of_guarantee_credit", period) block, then this PR's block, and return where(passported | on_income_related_benefit, 0, ...) in both files. This was resolved and run.
  • On main, every mixed-age HB recipient on income-related ESA already has nil applicable income, but only because main's Guarantee Credit test (any_over_SP_age & guarantee_credit > 0) reads a guarantee credit computed for families who cannot get Pension Credit. So this PR changes no mixed-age family against main.
  • Key the Housing Benefit Guarantee Credit passport on receipt #1927 alone lowers Housing Benefit for some of those couples (none to zero). They are too few survey records to report separately. With both PRs merged, every mixed-age couple on income-related ESA has the same HB as on main (0 records differ), and this PR adds +£18.3m against Key the Housing Benefit Guarantee Credit passport on receipt #1927's branch.
  • Key the Housing Benefit Guarantee Credit passport on receipt #1927's other mixed-age losers are not on these benefits. That is Key the Housing Benefit Guarantee Credit passport on receipt #1927's intended effect.
  • Replace generic child and adult flags with each programme's legal definitions #1896 has the same effect on these couples as Key the Housing Benefit Guarantee Credit passport on receipt #1927, by making guarantee_credit zero outside Pension Credit eligibility (see Impact).

• Abolish working-age Housing Benefit from 1 July 2026 (NI 1 October 2026) #1910 (working-age HB abolished from 1 July 2026 in GB). Clean merge. On Abolish working-age Housing Benefit from 1 July 2026 (NI 1 October 2026) #1910's branch this PR adds +£4.01m in 2026 (awards are paid for the part of the year before abolition) and £0 in 2027.

• Same-line edits elsewhere:

  • Count savings credit in Housing Benefit income where the Pension Credit award is savings credit only #1945 and Treat income derived from capital as capital in the legacy means tests #1995 add documentation and reference to housing_benefit_applicable_income;
  • Count savings credit in Housing Benefit income where the Pension Credit award is savings credit only #1945 and Value means-tested capital less 10% sale expenses and secured debt (UC reg 49 and legacy equivalents) #1969 edit housing_benefit_assessable_capital;
  • Replace generic child and adult flags with each programme's legal definitions #1896 and Use the State Pension Credit qualifying age, and start-of-year status for Class 4 NI #1907 already conflict with main in these files.

  Each resolution keeps both sides.

Impact (Enhanced FRS 2024-25, real runs)

The base is main 2c94286, the main this head contains (it includes #1896, #1939, #1940, #1944 and #1950); the branch is this PR's head, 2febacb. The two trees differ only in this PR's own files. Re-run on 2026-10-03. The earlier run against main 3c48247 (head 678334c) gave the same changes, gainer counts and poverty figures to the precision shown; only the base totals and the 2027-28 mean gain (£2,048) differed. Runs against main 8b2e6e1 with #1896 merged in also gave the same changes.

	2025-26	2026-27	2027-28
Housing Benefit, base	£13.498bn	£14.219bn	£14.593bn
Change	+£18.2m	+£18.5m	+£17.4m
Benefit units gaining	10.5k	9.4k	8.5k
Losing	0	0	0
Mean gain per gaining unit	£1,730	£1,982	£2,051
Pension Credit, UC, CTR, benefit cap	unchanged	unchanged	unchanged
AHC poverty, all people	−0.0009pp	−0.0012pp	−0.0009pp
BHC poverty, all people	−0.0013pp	−0.0013pp	−0.0010pp

• Almost all the gain is to working-age benefit units. In 2026-27 every gaining unit is on income-related ESA; in 2025-26 most are (Income Support is most of the rest).
• 22.4k working-age HB recipients were on these benefits in 2026. 13.6k of them already had applicable income at or below their applicable amount, so they were at maximum. The other 8.7k had applicable income above it and gain; 0.4k of them also had tariff income from capital.
• A few mixed-age couples on income-related ESA also gain on this base. They are too few survey records to report separately. Replace generic child and adult flags with each programme's legal definitions #1896 makes guarantee_credit zero for a family that is not eligible for Pension Credit, so main's Guarantee Credit test no longer zeroes their income by accident. This PR's passport does it on the lawful ground (reg 5(1)(b)).
• The caseload is essentially unchanged: almost every gainer already received some Housing Benefit.
• Differential check: no benefit unit outside the passport changes (0 of 61,223 records, in each year).
• The change in household net income equals the Housing Benefit change in each year.
• Against main alone, without Replace generic child and adult flags with each programme's legal definitions #1896, the working-age effect is the whole effect:
  • at main 8b2e6e1: +£16.65m, +£17.10m and +£16.06m, for 9.9k, 8.7k and 7.9k units;
  • at the earlier main 7b9fc37, before Cap the eligible rent at the LHA before the Housing Benefit taper #1926 took the taper off the LHA-capped rent: +£16.3m, +£16.85m and +£15.8m. Some passported LHA tenants were then already at the cap; now the passport raises them too.

Invariants

Property-tested in test_housing_benefit_income_related_benefit_passport_properties.py: Hypothesis, derandomize=True, 5 examples of up to 20 families each. The families cover single, couple, lone parent, couple with children, mixed-age and pension-age shapes, all three renting tenures and optional non-dependants, with random earnings, pensions, savings and awards.

1. Receipt: true exactly when Income Support, income-based JSA or income-related ESA is positive.
2. Passport: a family in receipt has nil applicable income, assessable capital and tariff income, and meets the capital limit.
3. Maximum: every family's entitlement is at most maximum HB (rent, capped at the LHA rate for LHA tenants, less non-dependant deductions, floored at zero). A family in receipt gets exactly that.
4. Metamorphic, income and capital: for a family in receipt, raising earnings, pensions and savings by the same amount changes neither entitlement nor eligibility.
5. Metamorphic, receipt: adding a positive Income Support, income-based JSA or income-related ESA amount never lowers HB before the benefit cap, and never removes eligibility. After the cap it can, because those benefits count towards the cap. That is an intended difference, so the test reads housing_benefit_pre_benefit_cap.
6. Ages: for social tenants with no non-dependants, a passported couple's entitlement is the rent whether its members are 40 and 35, 70 and 35, or 70 and 68.
7. Contributory benefits: contribution-based JSA and contributory ESA never set the passport, and the family's savings remain its capital.

Plus the dataset-wide differential check above.

Mutation check. Seven mutations were each applied to a scratch copy of this branch (b6d45a5). Each was run against the new and changed YAML files and the property tests, with Hypothesis shrinking off. The unmutated control passes both. The YAML and the property tests each catch all seven:

• M1, no income passport;
• M2, no capital passport;
• M3, the passport applied only where no one is over State Pension age (as Take Housing Benefit earnings disregards from net earnings only #1908's para 12 does);
• M4, receipt ignores income-related ESA;
• M5, receipt ignores Income Support;
• M6, receipt also counts contribution-based JSA and contributory ESA;
• M7, receipt needs more than £100 a year of a benefit instead of any amount.

Tests

• New: housing_benefit_income_related_benefit_passport.yaml, with the seven cases above. A to E fail on main. The controls F and G have the same financial results on main; their files also ask for the new receipt variable, which main does not have.
• New: in_receipt_of_income_support_jsa_ib_or_esa_ir.yaml, seven cases. They include a reported income-based JSA award screened out by the capital test, which is not receipt.
• New: the property tests (about 70 s).
• Run locally on 678334c: all 634 YAML tests under finance/benefit/family, gov/dwp and gov/local_authorities pass (including the new files), and so do the 1,750 tests/code_health tests and the property tests. CI runs the full suite.

Independent reviews (Opus 5.5 on subfleet; files in ~/reviews/uk-hb-ir-passport-2026-10-01/out/):

• r1, APPROVE at b6d45a5. It reproduced the 2026 impact on the then-current base exactly (+£16.85m, 8.47k gainers, 0 losers) and the interplay figures. Its minor findings were test and text hygiene. The two test fixes were in Tighten the Housing Benefit passport tests (follow-up to #2000's review) #2048 (tests only), now merged into this branch. The text fixes are applied here and in rulespec-uk#375.
• r4, at 678334c. It found the code approvable: the diff is exactly this PR's own files, the mechanism holds on today's main, and CI passed on that head (1,625 YAML and 2,261 pytest). One privacy blocker: a subgroup count in this body rested on fewer than 10 survey records. That count has been removed. It also recommended folding Tighten the Housing Benefit passport tests (follow-up to #2000's review) #2048's property-test fix into this branch before merging. Main now has Date the mixed-age couple Pension Credit exclusion and model the SI 2019/37 saving #1940's mixed-age Pension Credit saving, which can give a pension-age family in that control a guarantee credit and so nil capital.
• Head ec1a87c = 678334c + Tighten the Housing Benefit passport tests (follow-up to #2000's review) #2048 (b6d34c8) + a clean merge of main bfc5fea. On it, the HB YAML directory and the receipt file (103 tests) and the 5 property tests pass locally.
• r5, APPROVE at ec1a87c (subfleet 20261003-090934-hb2000-review-r5-delta). The delta since r4 is exactly Tighten the Housing Benefit passport tests (follow-up to #2000's review) #2048 plus a conflict-free main merge. Tighten the Housing Benefit passport tests (follow-up to #2000's review) #2048's changes are correct, and privacy passes: the smallest figure in this body rests on 14 survey records.
• Head 2febacb = ec1a87c + a clean merge of main 2c94286 (Cut PR CI from hours to about 20 minutes: Linux arm64, 4 xdist workers, cancel superseded runs #2115, Linux CI only). This PR's 10 files are byte-identical to ec1a87c. CI passed on this head (all 7 checks, runs 37127454596 and 37127454571).
• r6, APPROVE at 2febacb (subfleet 20261003-094843-hb2000-review-r6-delta). The merge is automatic and Cut PR CI from hours to about 20 minutes: Linux arm64, 4 xdist workers, cancel superseded runs #2115 does not change how this PR's tests run.

Research files (law texts, impact harness, scratch-merge runs and aggregate diagnostics) are in ~/reviews/uk-hb-ir-passport-2026-10-01/ on the author's machine. No record-level survey values are reported here.

axiom: uk/regulation/uksi/2006/213/schedule/5/paragraph/4, uk/regulation/uksi/2006/213/schedule/6/paragraph/5 TheAxiomFoundation/rulespec-uk#375 queued (extended with a dispatch-ready section: TheAxiomFoundation/rulespec-uk#375 (comment))

🤖 Generated with Claude Code

[MaxGhenis]

Merged at exact head 2febacb (gh pr merge --merge --match-head-commit 2febacb771cff99d9f8d89cfc610d72d36270d40); merge commit 0eccaaa.

Gates checked immediately before merging:

• CI: gh pr checks 2000 exited 0 on head 2febacb. All 7 checks passed: Lint, Test, Test documentation builds, and Install + smoke-import for py3.11 to py3.14 (runs 37127454596 and 37127454571).
• Mergeable: GitHub reported MERGEABLE and CLEAN. The PR was not a draft and had no CHANGES_REQUESTED review.
• Independent review of this exact head: r6 APPROVE at 2febacb (Opus 5.5, subfleet job 20261003-094843-hb2000-review-r6-delta, file ~/reviews/uk-hb-ir-passport-2026-10-01/out/review-r6-delta.md). It builds on r5 APPROVE at ec1a87c (20261003-090934-hb2000-review-r5-delta), r4 at 678334c and r1 APPROVE at b6d45a5 (20261001-113806-hb2000-review-r1). 2febacb is ec1a87c plus an automatic merge of main 2c94286 (Cut PR CI from hours to about 20 minutes: Linux arm64, 4 xdist workers, cancel superseded runs #2115, CI only). This PR's files are byte-identical to the ones r5 approved.
• Stack order: Replace generic child and adult flags with each programme's legal definitions #1896 merged first (3525263, 2026-10-02), as the PE-UK merge train set. Main moved after 2c94286 only by Encode fuel duty as its dated statutory schedule #1883 (fuel duty), which touches no DWP file.
• Impact stated: the Enhanced FRS impact was re-run today against main 2c94286, the main this head contains, with the head as the branch. The results are +£18.2m, +£18.5m and +£17.4m of Housing Benefit in 2025-26, 2026-27 and 2027-28, with 0 losers and no benefit unit outside the passport changed. The figures match the earlier run; the body's base totals and the 2027-28 mean gain were updated.

The merge follows Max's 2026-10-02 merge-on-gates rule for PE-UK statutory fixes (decision d786): CI green on the head, an independent review approving that exact head, the impact stated in the PR, and stack order allowing it.