Bind the Lifetime ISA holdings that microcosm datasets now carry, and count them as means-tested capital #1981

What the datasets carry

PolicyEngine/microcosm#1059 (for PolicyEngine/microcosm#1003, merged 1 October 2026) adds a was_lisa stage to the UK spine. UK datasets built by microcosm now carry three Lifetime ISA (LISA) cells, imputed from the Wealth and Assets Survey (WAS) round 8 person file:

• person.has_lifetime_isa (bool): the person holds one or more LISAs.
• person.lifetime_isa_balance (float, £): the current value across the person's LISAs. It is zero for non-holders and under-18s.
• household.household_lifetime_isa_balance (float, £): the sum over the household's persons.

They ship as allowed extra export columns. No policyengine-uk variable declares them, so the engine ignores them today.

Proposed variables

• lifetime_isa_balance
  • Person, YEAR, float, GBP, quantity_type = STOCK, an input.
  • Uprated by gov.economic_assumptions.indices.obr.per_capita.gdp, like savings and gross_financial_wealth.
  • The documentation should state the accounting in the next section.
• has_lifetime_isa
  • Person, YEAR, bool.
  • Formula: lifetime_isa_balance > 0. The dataset's stored column agrees with it by construction.
• household_lifetime_isa_balance
  • Household, YEAR, float, GBP, STOCK.
  • adds = ["lifetime_isa_balance"].

Wealth accounting

WAS counts every ISA, LISAs included, inside household gross financial wealth (HFINWR8_SUM, the engine's gross_financial_wealth). The LISA is therefore already a component of gross_financial_wealth and net_financial_wealth, and adding it to either would double count it. It is not part of savings (savings accounts) or corporate_wealth (shares, including stocks-and-shares ISAs), and must not be folded into either.

total_wealth sums property_wealth, corporate_wealth, private_pension_wealth and savings. It therefore omits LISAs today, as it omits cash ISAs, which the datasets export as household.cash_isa with no engine variable.

Adding household_lifetime_isa_balance, and a bound cash ISA, to total_wealth would not double count. It would move total_wealth, net_wealth and the wealth deciles, so it is a decision to take here rather than a side effect.

Means-tested capital

• The rule. The six means tests below value capital at its "current market or surrender value":

  • Universal Credit Regulations 2013 reg. 49(1);
  • Income Support Regulations 1987 reg. 49;
  • Jobseeker's Allowance Regulations 1996 reg. 111;
  • Employment and Support Allowance Regulations 2008 reg. 113;
  • State Pension Credit Regulations 2002 reg. 19;
  • Housing Benefit Regulations 2006 reg. 47, and reg. 45 of the pension-age Housing Benefit Regulations 2006.

• DWP's guidance. DWP applies that rule to the LISA in the same words for each benefit it administers. The LISA counts as capital from the outset at 75% of its value when the person is under 60, and at 100% from 60:

  • ADM H1659 for Universal Credit;
  • DMG 29665 for Income Support and income-based JSA;
  • DMG 52665 for income-related ESA;
  • DMG 84751 for Pension Credit.

• Universal Credit, confirmed. The government confirmed the Universal Credit treatment in its response to the Treasury Committee's Lifetime ISA report (Third Special Report of Session 2024–26, 11 September 2025, response to recommendation 3) and in written answer UIN 46426 (23 April 2025).

• Housing Benefit. Local authorities administer it, and its guidance manual's capital chapter predates the LISA. Its regulations use the same valuation words.

• The charge. It is set in the Schedule (Lifetime ISA further provision) to the ISA Regulations 1998:

  • Paragraph 5 sets 25%.
  • Paragraph 5A, inserted by SI 2020/506, set 20% for withdrawals from 6 March 2020 to 5 April 2021. DWP's guidance quotes 25% only.
  • Withdrawals carry no charge from age 60 (paragraph 4(1)), with a terminal illness (paragraph 4(2)), or towards a first home (paragraph 6(2)).

• The product's future. HM Treasury's First Time Buyer ISA consultation (22 June to 18 August 2026) proposes a product offered in place of the LISA once it is available. Until then, LISAs can still be opened, and holders can keep saving into them under the existing rules indefinitely (paragraph 1.1). The charge therefore stays for existing LISAs.

• The gap. The six capital.sources lists name household stocks only: savings, property, land and corporate_wealth. So every means test understates LISA holders' capital.

• Attribution. Those stocks are known only per household, so each regime allocates them to benefit units by a proxy:

  • an adult share for Universal Credit and Housing Benefit;
  • the claiming benefit unit for Income Support, income-based JSA and income-related ESA;
  • the pension-age adults for Pension Credit.

  A LISA is one person's holding, and most holders are young adults, many of them living with their parents. Through the proxies, an adult child's LISA would count against a parent's income-based JSA claim, or a grandchild's against a grandparent's Pension Credit. The LISA should count only in its holder's benefit unit.

Proposed

• Parameters:
  • gov.hmrc.isa.lifetime.withdrawal_charge: 0.25 from 2017-04-06, 0.20 from 2020-03-06, and 0.25 again from 2021-04-06;
  • gov.hmrc.isa.lifetime.charge_free_age: 60.
• A person variable lifetime_isa_countable_capital:
  • lifetime_isa_balance × (1 − withdrawal_charge) below the charge-free age, and the full balance from it;
  • terminal illness and the first-home route are not modelled.
• In each of the six means tests, count it as capital of the holder's benefit unit, through a list of person-level capital sources beside the household sources.
• Tests:
  • a Universal Credit claimant under 60 with a £10,000 LISA counts £7,500;
  • the same claimant at 60 counts £10,000;
  • in 2020-21 the claimant counts £8,000;
  • an adult child's LISA counts in the child's benefit unit and not in the parents'.
• Cash ISAs are countable capital too and are missing from every list; cash_isa is exported per household. Binding it is a separate change with a much larger effect.

Data caveats

These come from the microcosm receipts, experiments/1003-uk-lisa-receipts.md.

• Donor and vintage. Balances are April 2020 to March 2022 pounds from a Great Britain donor. microcosm does not uprate them, and Northern Ireland adults are predicted from the Great Britain model.
• Ownership level.
  • On the candidate spine, 1.25% of adults hold a LISA, against 0.89% in the WAS donor.
  • HMRC counted 1.14 million LISA subscribers in 2024-25 (provisional, ISA statistics Table 9.4); the candidate spine has about 660,000 holders, and WAS itself undercounts against HMRC.
  • The ownership level is not recalibrated to HMRC; the options are set out in the receipts, Part F.
• Holdings only. There are no contribution, bonus or withdrawal fields. Reforms to the annual limit, the bonus rate, the withdrawal charge's revenue or the first-home route cannot be scored from these holdings.

Done when

• The three variables are declared, documented and tested.
• The LISA's realisable value counts as capital of the holder's benefit unit in Universal Credit, Housing Benefit, Income Support, income-based JSA, income-related ESA and Pension Credit.
• The total_wealth decision is recorded.
• The effect on Universal Credit caseload and spending is measured on a microcosm dataset that carries the cells, and reported in the PR.