Stop counting interest, dividends and rent as Universal Credit unearned income

[MaxGhenis]

Summary

Universal Credit counts capital through its assumed yield (tariff income), not through the interest, dividends or rent it actually pays. PolicyEngine-UK counted savings_interest_income, dividend_income and property_income as UC unearned income. It subtracted them only when tariff income was positive and a household asset flag for that stream was set (capital_derived.yaml). So the model:

• counted all three in full when capital was £6,000 or less, or when the data recorded no matching asset; and
• added interest on top of tariff income from other capital whenever household savings was 0. In 2026, 304k UC-eligible benefit units had £2.73bn of interest counted this way. Only 35k of them had an award on main, so their UC rises by only £0.005bn.

This PR removes all three from gov.dwp.universal_credit.means_test.income_definitions.unearned from 2013-04-29. It also deletes capital_derived.yaml and the subtraction in uc_unearned_income, which becomes a plain sum of the listed sources.

Worked case (2026): a single claimant aged 30 with £5,000 of savings, £200 of interest, £300 of dividends and £1,000 of rent.

• On main: UC £3,598.80.
• On this branch: £5,098.80 (12 × £424.90). The capital is not above £6,000, so it yields nothing, and none of the actual income is unearned income.

Law

All texts were read from legislation.gov.uk (/data.xml, 2026-09-30). The verbatim copies are saved with the review notes.

UC Regs 2013 reg 66(1) (https://www.legislation.gov.uk/uksi/2013/376/regulation/66) defines a closed list:

(1) A person's unearned income is any of their income, including income the person is treated as having by virtue of regulation 74 (notional unearned income), falling within the following descriptions— … (k) income that is treated as the yield from a person's capital by virtue of regulation 72; (l) capital that is treated as income by virtue of regulation 46(3) or (4); (la) PPF periodic payments; (m) income that does not fall within sub-paragraphs (a) to (la) and is taxable under Part 5 of the Income Tax (Trading and Other Income) Act 2005 (miscellaneous income).

Sub-paragraphs (a)–(j) cover retirement pension income, listed benefits, foreign benefits, spousal maintenance, foreign state pensions, student income, ETA/ENTSA payments, Sports Council awards, income-protection insurance, annuity income and trust income. None of them covers a claimant's own interest, dividends or rent.

Reg 72 (https://www.legislation.gov.uk/uksi/2013/376/regulation/72). It is unamended since it was made, and the as-made text matches the current text:

(1) A person's capital is to be treated as yielding a monthly income of £4.35 for each £250 in excess of £6,000 and £4.35 for any excess which is not a complete £250.
(2) Paragraph (1) does not apply where the capital is disregarded or the actual income from that capital is taken into account under regulation 66(1)(i) (income from an annuity) or (j) (income from a trust).
(3) Where a person's capital is treated as yielding income, any actual income derived from that capital, for example rental, interest or dividends, is to be treated as part of the person's capital from the day it is due to be paid to the person.

ITTOIA 2005: interest, dividends and rent are not Part 5 income, so reg 66(1)(m) cannot reach them.

• s.369(1): "Income tax is charged on interest." (Part 4, Chapter 2.)
• s.383(1): "Income tax is charged on dividends and other distributions of a UK resident company." (Part 4, Chapter 3.)
• s.402(1): "Income tax is charged on dividends of a non-UK resident company." (Part 4, Chapter 4.)
• s.268: "Income tax is charged on the profits of a property business." (Part 3.)
• s.575(2)–(3): income that could fall in Part 5 and also in Part 3 (UK property business) or Part 4 Chapters 2–3 "is dealt with under" Part 3 or Part 4.
• s.687(1), the residual Part 5 charge, applies only to "income from any source that is not charged to income tax under or as a result of any other provision of this Act or any other Act". That keeps out non-UK dividends and overseas rent, which s.575 does not list.

Reg 22(1) deducts only earned income (tapered) and "all of the claimant's unearned income". Welfare Reform Act 2012 s.8(3) allows deductions only for earned and unearned income "calculated in the prescribed manner". Income that is neither earned income (reg 52) nor within reg 66(1) is not deducted.

Northern Ireland. The UC Regs (NI) 2016 (SR 2016/216) regs 66(1) and 72 are the same on these points.

• NI reg 66(1)(m) reaches only ITTOIA Part 5, as GB's does.
• In both, (m) has been limited to Part 5 since it was made. The 2018 amendments only changed "(a) to (l)" to "(a) to (la)" (GB SI 2018/65 from 11.4.2018; NI from 8.5.2018). So the change starts at 2013-04-29.
• NI reg 72(3) omits the words "for example rental, interest or dividends" but has the same effect.

DWP Advice for Decision Making, Chapter H5 (unearned income) and Chapter H1 (capital), current GOV.UK versions:

• H5002 lists the reg 66(1) categories, then says: "If a type of income is not listed above, it does not affect the claimant's award."
• H5093: actual income derived from capital "has to be treated as capital from the date it is due to be paid to the claimant. It cannot be treated as income."
• H5094: the types of income derived from capital include "1. interest 2. dividends and 3. rental income". Example 2 covers a let second property: "the actual income derived from that capital (the rental income) is treated as part of Scoot’s capital from the date it is due to be paid".
• H5112: "Where a claimant receives an income which is not defined as such for the purposes of UC then it cannot be taken into account as income." Its example is a spare room let at £60 a week, which "is not taken into account as Teresa’s unearned income". Note 2: renting rooms "as part of conducting a trade, for example running a B&B" is taken into account as earnings.
• H1764: "Capital does not affect what benefit claimants can get if their capital is £6,000 or less."

What still counts, and how the model handles it

Route	Law	Model
Assumed yield on capital above £6,000	reg 72(1), reg 66(1)(k)	uc_tariff_income, unchanged
Letting that is a trade (B&B, guest house)	reg 52(a)(ii), reg 57; ADM H5112 note 2	Trading income sits in self_employment_income, which is already earned income. property_income is FRS subletting rent, "rent before tax from other property", and CVPAY (see Legal review).
Company the claimant effectively owns	reg 77. For a trading company, (3)(b) treats the company's income as self-employed earnings. For a property-business company, (2) treats the company's capital as the claimant's instead of the shares.	Not identifiable in the data. Dividends are not the measure in either the old or the new treatment. In progress for #1948.
Annuity income; trust income	reg 66(1)(i), (j)	Not among these three variables; unchanged
Part 5 income (royalties, settlor income, estates in administration)	reg 66(1)(m); ADM H5111	Not in these three variables. FRS royalties and trust income sit in miscellaneous_income, which UC wrongly counts as earned income (#1961).
Rent from property that is itself disregarded capital (for sale, business asset, a relative's home), or from a lodger	reg 66(1) is closed; ADM H5112	Not income. Legacy benefits count some of this (IS Regs 1987 reg 48(4) and Sch 9 para 22). UC has no equivalent provision.

Accrued yield. The model still does not add retained interest, dividends or rent to uc_assessable_capital under reg 72(3). The data record each capital stock at one point and do not say whether a year's receipts were retained, spent or are already in that stock. This was the reason capital_derived.yaml gave, and it is unchanged.

Legal review

Three adversarial reviews tried to refute the reading from three angles: the statutory text, DWP practice, and edge cases (trusts, annuities, estates, reg 77, lettings, Part 5). None found a route by which a claimant's own interest, dividends or rent is UC unearned income.

The memo, with verbatim texts, is kept with the review notes.

Refinements, all reflected in the parameter description:

• The exclusion rests on reg 66(1) being a closed list, not on reg 72(3). Reg 72(3) applies only "Where a person's capital is treated as yielding income". For capital of £6,000 or less, or disregarded capital, the receipts are simply outside reg 66(1).

• Some income derived from capital still counts:

  • as a trust beneficiary (66(1)(j));
  • as annuity income (66(1)(i));
  • as estate or other Part 5 income (66(1)(m)).

  None of it is in the three variables. From the FRS 2024-25 data dictionary (UKDA 9563) and uk-data datasets/frs.py:

  • savings_interest_income is interest on current, NS&I, savings, ISA, basic and credit-union accounts;
  • dividend_income is gilts, unit and investment trusts, and shares;
  • property_income is SUBRENT ("Amount of rent from subletting"), ROYYR1 ("Rent before tax from other property") and CVPAY ("Amount of rent after state benefits (boarders/lodgers)"). CVPAY is rent a boarder or lodger pays, which the build wrongly counts as their income; Stop counting rent paid by boarders and lodgers as their property income policyengine-uk-data#503 removes it. Royalties (ROYYR2) are in miscellaneous_income.

  ADM H1673 values unit trusts as the claimant's own capital.

• The literal reg 46(3) → 66(1)(l) loop. Could regular rent or interest that 72(3) deems capital come back as "paid regularly and by reference to a period"? The loop would make 72(3) pointless, and ADM H5093 says such income "cannot be treated as income". So the specific rule prevails. No case law was found.

The one genuinely unsettled point is on the earned side: is a landlord's letting a trade?

• The UC Regs do not define "trade".
• DWP's Treatment of earnings guidance V17.0 (deposited paper DEP2026-0403) says a self-employed landlord's "rental income is treated as self-employed income".
• For tax, passive letting is not a trade.
• In SSWP v Faghy (UA-2026-000266-USTA, UT, 27 August 2026), the Secretary of State had refused a holiday-let landlord gainful self-employment. The FtT found it, and the UT remitted on reg 77 and Sch 10 para 7 grounds without deciding the principle.

Rent is not unearned income under any reading, so the list change here is right either way. Under DWP's reading, though, some property_income should also count as self-employed earnings: 55% taper, and the minimum income floor if it is the main employment. Neither main nor this PR does that, and the data do not identify such landlords.

Main's full deduction of rent was no stand-in for that route either. It also deducted subletting rent in the home, which H5112 excludes in terms, and rent that lodgers pay (CVPAY), which is not income at all. And it removed rent entirely for £6,000–£16,000 of let property.

The independent review judged that this earned-side question does not make the change in this PR ambiguous.

Invariants (stated and tested)

For every input:

1. Single count. uc_unearned_income equals uc_tariff_income plus the other sources on the model's reg 66(1) list for the year, read from the parameter, and none of the three capital incomes is on that list. The test checks uc_tariff_income against an independent reg 72(1) calculation on uc_assessable_capital.
2. Invariance. Scaling interest, dividends and rent (×0, ×0.5, ×3) changes nothing in unearned income, tariff income, assessable capital, eligibility or the maximum amount. For families without earnings it changes nothing in the award either.
3. Monotone in capital. Adding capital to any countable source never raises the award, before or after the benefit cap. It never lowers tariff income, and it never makes an ineligible family eligible.

policyengine_uk/tests/test_uc_income_from_capital_properties.py runs these with Hypothesis over populations of up to 20 families in 2020 and 2026, across five tenures and five regions including Northern Ireland, with optional reported capital. All three fail on main.

Invariant 3's counterexample on main, reproduced by hand: a single claimant aged 30 with £3,000 of interest.

• With £6,000 of savings, UC is £2,098.80.
• With £6,001, UC is £5,046.60: one more pound of savings raises it by £2,947.80. The pound creates £52.20 of tariff income, which switched the interest out of unearned income.
• On this branch the same pound lowers UC from £5,098.80 to £5,046.60, exactly the tariff income. Both cases are pinned in uc_income_from_capital.yaml.

On this branch the properties pass: 4 passed and 1 strict xfail with 10 examples each, and the same with 60 examples each.

The award clause of invariant 2 is limited to families without earnings. The model deducts the whole benefit unit's income tax from earnings, so tax on dividends raises a working family's award (#1942). A strict xfail pins that case.

Tests

• uc_unearned_income.yaml: every case is rewritten with its legal basis.
  • Five expected values change, all on inputs where main counted the income:
    • "All sources summed" becomes 11,100, as interest and dividends drop out;
    • dividends without share wealth in the data;
    • rent without property capital;
    • rent with only a home;
    • dividends with only disregarded pension wealth.
  • New cases: all three streams below the £6,000 threshold; interest not added to tariff income from a let property when savings are 0.
• New uc_income_from_capital.yaml: seven award cases hand-computed from regs 22, 66 and 72 at 2026-27 rates (the arithmetic is in the file):
  • single claimant below the threshold: £5,098.80;
  • £10,000 of capital in let property, with rent and interest: £4,263.60, the same in Northern Ireland. The property value is taken as capital as entered. The model does not yet apply reg 49's 10% sale-cost and secured-debt deductions (UC capital ignores reg 49 valuation (10% sale costs and secured debts) #1960), and the test comments say so;
  • lodger in the home: £5,098.80;
  • working couple with a partner's dividends: £2,503.64;
  • the £6,000 / £6,001 savings pair above.
• The Hypothesis properties above.

Enhanced FRS impact

Both sides are real Microsimulation runs on private copies of enhanced_frs_2024_25.h5 (sha256 e433e532…), with no modified tracked files:

• main at 7b9fc37, which includes Count contributory ESA, Maternity Allowance, industrial injuries benefit and capped Carer Support Payment as UC unearned income #1958;
• this branch at b501021, rebased on it. The head has since been rebased onto later main by María Juaristi (b2369e4), and Count State Pension as Universal Credit unearned income #1943's property-income test was inverted. The model change is unchanged.

UK totals:

Year	UC main (£bn)	UC branch (£bn)	Change (£bn)	Gainers (k benefit units)	Newly entitled (k)	UC losers	AHC poverty (pp)	Child AHC poverty (pp)
2025	73.38	74.98	+1.60	1,714	97	0	−0.09	−0.23
2026	77.37	79.01	+1.64	1,682	99	0	−0.10	−0.28
2027	79.14	80.83	+1.69	1,688	104	0	−0.10	−0.28
2028	80.63	82.37	+1.74	1,683	104	0	−0.10	−0.28
2029	81.70	83.49	+1.80	1,675	103	0	−0.10	−0.28
2030	78.58	80.41	+1.83	1,660	107	0	−0.17	−0.50

The poverty changes rest on few survey records, like the aggregate. The 2030 jump in the poverty effect, with little change in the UC figure, is one sign of that.

The breakdowns below were measured before #1958 merged: main 44240bd against this branch at 191bfa0, on the same dataset. There the 2026 change was +£1.66bn with 1.70m gainers, so the totals barely moved on rebase. The model change is the same; only the base differs.

2026 breakdown:

• 1.70m benefit units gain, on average £977. Nobody loses UC.
• 103k units become newly entitled, taking £0.61bn of the gain.
• On main, UC counted £12.7bn of these three streams in 3.73m UC-eligible units, £1.08bn of it in units with an award.
• £1.62bn of the gain goes to units that had no tariff income on main (capital of £6,000 or less, or none recorded).
• The award before the benefit cap rises by £1.75bn. The award is max(pre-cap − (benefit cap reduction + uc_deductions), 0): the cap reduction takes £0.04bn, and deductions and the zero floor take the remaining £0.05bn.
• Council tax reduction moves by less than £0.1m.
  • The legacy-style local schemes use UC earned plus unearned income plus the award, so there a fall in unearned income offset by a rise in UC nets out.
  • The national schemes count universal_credit in applicable income, so part of the UC gain is withdrawn at the CTR taper, and newly entitled units change route.
  • About 5k benefit units lose a little CTR, and 0.2k households' net income falls.

Most of the £1.66bn rests on imputed dividends and missing capital

The rule change is right whatever the data. The size of the impact is not the effect of the law alone.

Dividends are imputed. policyengine-uk-data imputations/income.py impute_income() calls impute_over_incomes(dataset, model, ["dividend_income"]) on the FRS half too, so every respondent's dividends are an SPI QRF draw on age, gender and region (found on #1948; read in uk-data main b45c373). Splitting each gainer's gain pro rata by the gross interest, dividends and rent it records, for 2026:

Stream	FRS half (£bn)	SPI-synthetic half (£bn)	Total (£bn)
Dividends (imputed in both halves)	1.09	0.16	1.25
Rent	0.28	0.04	0.32
Interest	0.06	0.04	0.10
All	1.43	0.23	1.66

The table's attribution is a convention wherever the award, not the counted income, limited the gain. As a real-run subset: FRS-half gainers with interest or rent and no dividends gain £0.32bn, and there the whole gain comes from reported interest and rent.

Real rebuilds with the dividend fix. The #1948 session rebuilt the Enhanced FRS twice at production settings (512 epochs):

• from uk-data main (b45c373), dataset sha256 6baf2953…;
• from Keep FRS-reported dividends and key them on person_id policyengine-uk-data#498 (c9b203e), which keys FRS dividends on person_id and stops overwriting them, dataset sha256 c9878b7f….

Both were run with policyengine-uk main 44240bd and with this PR at 473adff (the same capital-income rule change, measured before the rebase onto #1958). UC change from this PR, £bn:

Year	uk-data main rebuild	uk-data #498 rebuild
2025	+1.88	+0.85
2026	+1.93	+0.90
2027	+2.00	+0.92
2028	+2.04	+0.91
2029	+2.10	+0.93
2030	+2.13	+0.94

With dividends that are reported rather than imputed, the change is about £0.9bn a year. That is roughly half of what it costs on today's data. The rebuild of main differs from the published artifact (+£1.66bn in 2026), because of rebuild settings.

Capital stocks and income flows disagree.

• £0.95bn of the £1.30bn gain in units with dividends is in units with no corporate_wealth, and those dividends are imputed.
• £0.34bn of the £0.38bn gain in units with interest is in units with no savings.
• All the gain in units with rent is in units with no property other than the home. That partly follows from selection: units with £6k–£16k of let property already had rent removed on main, and units above £16k are ineligible in both runs.
• Part of property_income has no let property behind it: subletting rent in the claimant's own home, which the law excludes, and CVPAY, which is rent paid and not income (Stop counting rent paid by boarders and lodgers as their property income policyengine-uk-data#503). That part of the rent gain is not missing capital. The data do not separate the components on the Enhanced FRS.

Real-run sensitivity. Both main and this branch were run on a copy with capital raised to income ÷ yield (4% interest and dividends, 7% rent).

• On that copy the change is worth £0.04bn.
• The same top-up cuts main's UC from £79.3bn to £71.3bn and its caseload from 6.35m to 5.78m, well below DWP counts.
• It builds capital from imputed dividends, subletting rent and CVPAY, so it is not a sound bound either.

Concentration. The 10 records with the largest weighted gain carry 46% of it, and the 100 largest carry 85% (1,284 gaining records).

The data side is PolicyEngine/policyengine-uk-data#495, covering capital for reported interest and rent, and the #1948 work on keeping FRS-reported dividends and adding director dividends. PolicyEngine/policyengine-uk-data#498 fixes the dividend part. Until the data fixes land, quote this change's effect on households, not its aggregate.

Coordination

• Count State Pension as Universal Credit unearned income #1943 adds state_pension to the same list. Its property test test_state_pension_counts_like_property_income asserts that UC treats property_income held without property capital like State Pension. That is no longer true on this branch. Whichever of the two merges second must drop that equivalence or change it to "property income does not count". The list entries themselves do not conflict.
• Exclude corporate_wealth from Universal Credit countable capital pending a pension split #1837 edits capital_derived.yaml, which this PR deletes.
• UC earned income deducts the whole benefit unit's tax, including tax on pensions and other unearned income #1942 / Deduct only each person's own tax and NI on earnings from UC earned income #1949: the whole-unit tax deduction, pinned by the xfail above. Deduct only each person's own tax and NI on earnings from UC earned income #1949 fixes it. Whichever of Deduct only each person's own tax and NI on earnings from UC earned income #1949 and this PR merges second removes the xfail marker and re-runs the combined impact.
• UC-eligible units with six-figure dividends: dividends treated as yield of a few thousand pounds of imputed capital #1948 (UC units with six-figure dividends): this PR does not resolve it. Those units already had their dividends dropped on main. This PR answers its legal question: dividends are never UC unearned income. The fixes are the dividend imputation (data) and reg 77, which is in progress on branch uc-dividends-owner-managers-1948.
• Count contributory ESA, Maternity Allowance, industrial injuries benefit and capped Carer Support Payment as UC unearned income #1958 (reg 66(1)(b) benefits) merged first. It split the list into 2013-04-29 and 2023-11-19 blocks, each still naming the three capital incomes. Git rebased this branch without a conflict but removed them from the later block only, so this PR removes them from both by hand (b501021) and keeps Count contributory ESA, Maternity Allowance, industrial injuries benefit and capped Carer Support Payment as UC unearned income #1958's entries and references. The single-count property now reads the list from the parameter and asserts that none of the three is on it.
• UC capital ignores reg 49 valuation (10% sale costs and secured debts) #1960 / branch uc-capital-sale-expenses (reg 49's 10% sale-cost deduction) touches none of this PR's files.
  • On main alone, that deduction takes some units to £6,000 or less of capital, where main counts their interest, dividends and rent again. That session measured this as about £132m of UC lost in 2026.
  • This PR removes that cliff; the reg 49 PR will report the stacked effect.
  • After it lands, property values in the UC tests here are valued at 90%, and the test notes saying "reg 49 is UC capital ignores reg 49 valuation (10% sale costs and secured debts) #1960" should be updated.

axiom: uk/regulation/uksi/2013/376/66 encoded-correct (TheAxiomFoundation/rulespec-uk#388, merged first: 66.test.yaml cases bank_interest_is_not_unearned_income, dividends_are_not_unearned_income, rent_from_let_property_is_not_unearned_income, rent_from_a_lodger_in_the_home_is_not_unearned_income, assumed_yield_on_let_property_counts_as_unearned_income. #388 also pins reg 72(3) in 72.test.yaml, where Axiom is ahead of PolicyEngine, which does not model reg 72(3) accrual.)

Follow-ups

• Capital stocks disagree with income from capital, so UC capital tests miss rent, share and savings holders policyengine-uk-data#495: capital stocks inconsistent with reported interest and rent; dividend imputation.
• Reg 77 owner-managed companies: in progress for UC-eligible units with six-figure dividends: dividends treated as yield of a few thousand pounds of imputed capital #1948.
• UC capital ignores reg 49 valuation (10% sale costs and secured debts) #1960: UC capital ignores reg 49 valuation (10% sale costs and secured debts).
• UC counts royalties, trust income and overseas pensions in miscellaneous_income as earned income #1961: UC counts royalties, trust income and overseas pensions in miscellaneous_income as earned income.
• Legacy benefits (HB, IS) still count property_income. IS Regs 1987 reg 48(4) treats income from capital as capital, except income from certain disregarded capital (Sch 9 para 22). A separate session will verify and fix HB and IS.
• Letting as earned income: through reg 52(a)(ii) and 57 (trade) or reg 52(a)(iii) ("any other paid work"). Model it if the data can identify landlords whose letting is their business (see Legal review).

🤖 Generated with Claude Code

[juaristi22]

Codex review — ready to merge

Reviewed commit b501021b277b704ca4dca5fbc4e41473c9ff1591. COMPLETE; 0 confirmed blocking finding(s). Independent code/test and policy/source reviews were consolidated for the changed behavior and affected dependencies.

No actionable introduced code/test defect found. Both dated unearned-income lists remove the three actual capital-income variables, and uc_unearned_income becomes a direct aggregation of the remaining list including tariff income. The obsolete conditional subtraction and its parameter are removed together, avoiding a double subtraction or dangling runtime reference. The implementation and tests distinguish actual property-income flows from the separate capital-value/tariff calculation.

The strict expected failure for dividend tax reducing UC earned income identifies the unchanged whole-benefit-unit tax allocation in#1942. That formula is outside this diff; the issue is exposed by this correction and should remain an integration follow-up, not be misreported as a new formula defect here. Coordinate the dated-list changes with#1943 and update its private-pension/property-income equivalence test and documentation when the branches are combined. Retained-capital timing and the separately documented letting-as-trade classification remain scope assumptions. Confidence: high for code/test behavior in the stated scope.

No confirmed policy defect in removing actual interest, dividends and ordinary property rent from the UC unearned-income source list.

UC reg66 uses a closed list of income categories; actual savings interest, dividends and ordinary rental income are not included merely because they are taxable or because capital is below£6,000. Reg72 counts assumed yield from qualifying capital and treats its actual receipts as capital, subject to the annuity/trust exceptions. Removing the old conditional capital-derived subtraction therefore fixes the zero-tariff/no-matching-asset cases without requiring actual income to be re-added. The current diff removes the three variables from both dated lists and retains uc_tariff_income; the stale conditional subtraction/unused parameter are removed together.

Sources: UC66, UC72, NI equivalent closed list.

Trading/owner-manager earnings, trust/annuity income and retained receipts as capital are distinct categories. The PR documents the landlord/trade representation limit; it does not warrant moving every company or trading receipt into these removed ordinary-income variables. Existing tax-on-unearned-income allocation defects belong to#1949, not this diff. Resolve shared-list edits with#1943 by retaining its State Pension addition. No evidence-backed new regression was found.

Validation: 26 passed, 1 xfailed, 1 warning. Changed implementation and relevant final-award/dependency tests were checked at this exact commit. 3 official source originals were inspected and their cached bytes verified. No material review gaps remain within this scope; private population-impact claims were not independently rerun.

Focused tests used an existing cached Python 3.13 environment (Core 3.32.9, NumPy 2.1.3, pandas 2.3.1, microdf-python 1.2.1, Hypothesis 6.168.2, pytest 8.4.2) with this PR’s isolated source snapshot. No dependencies were installed; this was not a freshly synced lock environment or a full-suite/population run. This is a review comment, not a formal GitHub review vote.

Live check before posting (2026-10-02T10:43:04.069390+00:00): same commit, CLEAN; all reported CI checks pass.

[juaristi22]

Rebased onto main 8b2e6e174c9ed1fd6ace8b1ca5467de272c31342 and pushed 3a6b1a2b5aef2e210c2d086b1e30be8b48b62cdd.

The combined changes exposed an obsolete test that treated State Pension and ordinary rental income as equivalent for Universal Credit. The repaired test checks that rental income leaves UC unchanged, while State Pension reduces the award before the benefit cap. It preserves private-pension equivalence and the existing tax-allocation expected failures. The UC notebook now describes the same capital-income treatment.

Validation on the rebased head: both affected Python files plus UC income YAML passed 72 tests, with 2 expected failures; the full UC YAML directory passed 174 tests. Ruff formatting, lint and whitespace checks passed. GitHub CI is pending.