Repository navigation
Count per-person cash and stocks-and-shares ISA balances in the holder's benefit unit #2729
Description
#2712 adds person inputs cash_isa_balance and stocks_and_shares_isa_balance, and makes the household cash_isa and stocks_and_shares_isa their sums.
The means tests still read the household totals. The capital sources for UC, Housing Benefit, Pension Credit, Income Support, JSA and ESA pool those totals with the household's other financial capital, and allocate the pool between benefit units. So a lodger's or an adult child's cash ISA is shared with the main family's capital.
#1983 solved this for the Lifetime ISA. lifetime_isa_countable_capital sits in each programme's person_sources, which count the claimant's and partner's holdings in their own benefit unit.
What's needed:
- Count
cash_isa_balancethroughperson_sourceswhere person balances are given. - Do the same for
stocks_and_shares_isa_balance. It sits insideshare_like_capitalandcorporate_wealth, so it must not be counted twice. - Keep the household proxy for datasets that store only household totals, as all current Microcosm exports do.
- Tests: a lodger's cash ISA stays out of the main benefit unit's capital, and household-only data gives today's results.
Raised in Pavel's review of #2712 (nice to have 10). It needs #2712 merged. For population runs it also needs the person balances from PolicyEngine/microcosm#1415.
Activity
vahid-ahmadi commentedon Oct 29, 2026
Checked against main at 2c7bf5a (after #2712, merged at 2c5e297). Code reading only; I ran nothing.
The claims hold.
cash_isaandstocks_and_shares_isaare household sums of the person balances (variables/household/wealth/cash_isa.py:19,stocks_and_shares_isa.py:21). All six programmes carrycash_isaandshare_like_capitalinsources(for exampleuniversal_credit/means_test/capital/sources.yaml:49-51,pension_credit/income/capital/sources.yaml:30-35), and onlylifetime_isa_countable_capitalinperson_sources. The pool is shared by claimant and partner counts (uc_assessable_capital.py:43-69), so a lodger's cash ISA does reach the main family today.should: cash has the same double-count risk the issue names for stocks and shares.
cash_isakeeps a stored column (cash_isa.py:9-12), so a dataset can hold both a household total and person balances. Ifcash_isa_balancegoes intoperson_sourceswhilecash_isastays insources, those balances count twice. Suggest the pool takesmax(0, cash_isa - household sum of cash_isa_balance)andperson_sourcestakescash_isa_balance. Household-only data then gives today's result exactly, person-only data puts nothing in the pool, and there is no need to detect which kind of data you have.should: for stocks and shares, net the person sum out of
share_like_capital, notcorporate_wealth, becauseshare_like_capitalalready chooses betweencorporate_wealthand the itemised sum (share_like_capital.py:55). One test wherecorporate_wealthexceeds the components and one with components only would pin both branches.should: add a test with a dependent child's ISA balance. Today the pool hands it to the adults. Through
person_sourcesthe claimant/partner mask (uc_assessable_capital.py:47-51) drops it, which fits UC: the capital condition is the claimant's (Welfare Reform Act 2012 s.5). I did not check the legacy benefits' rules on children's capital. Also say incash_isa_balance's docs whether Junior ISAs are in it.🤖 Generated with Claude Code