Facts request from the microcosm side (microcosm#840 assessment; register items microcosm#736 items 9/13; deferral tracker microcosm#796). Follow-on to #187/#188 (the April–December 2025 family-type and number-of-children histories).
Why
The four UC with-children calibration cells (dwp.uc.households_single_with_children, …children_{1,2,5_or_more}) bind the Stat-Xplore Family Type and Number of Children splits, whose "with children" concept is not the one the model measures. Per the Stat-Xplore metadata:
- Family Type: from April 2019 "families with children are determined by a child or young person under 20 that has been verified as living in the household" (metadata).
- Number of Children: "all children or young people declared as living in a household on Universal Credit who are under 20 and whose details have been verified by DWP" — explicitly including over-16s in non-advanced education, looked-after children, and "young people in multigenerational households whose parents are not the claimant"; "it is currently not possible … to reliably identify if a declared child is a dependant or a child of the claimant" (metadata).
Measured on the current microcosm candidate (v18), UC households "with children" are 2.37m on the model's own-children concept but 3.06m on DWP's any-verified-under-20 concept, against DWP's 3.13m — the deficit the deferrals record is mostly a concept gap. The field that carries DWP's count on the concept the model computes (a child element in the award) is Child Entitlement (Yes/No): "whether a household receives a child element in their Universal Credit award … responsibility for a child or qualifying young person" (metadata).
What
Two new record sets from Stat-Xplore UC_Households (count measure V_F_UC_HOUSEHOLDS), Great Britain, monthly April–December 2025 (same months as #188's packages so the calendar-average operation is sum-consistent):
- Family Type × Child Entitlement — 5 family-type values × {Yes, No}.
- Number of Children × Child Entitlement — 0, 1, 2, 3, 4, 5 or more, unknown × {Yes, No}.
Optional third, if cheap in the same query: Family Type × Monthly Award Amount band "No payment" only, monthly — the nil-award households the model's universal_credit > 0 measure can never match (May 2025: 66,982 single-with-children, 73,849 couple-with-children, 501,957 in total; today only the May 2025 band package carries it — see the companion issue on band histories).
Shape
Follow #188's pattern: facts-only, publisher values untransformed, period typed monthly, benefit-unit entity, geography K03000001, assertion: observation, provenance to the archived API response (query echoed). Distinct source concepts or a child-entitlement dimension on the existing dwp.uc_households concept — author's call under the loader guarantees; the consumer register will re-author its family-type and children references to whichever shape lands.
Consumer
microcosm uk_population_targets.json: the four family-type and five number-of-children cells move to "with child entitlement" counts (calendar-year average), which the engine measures directly (uc_child_element > 0 / the num_uc_children variable); the nil-award band becomes a value declaration on the same cells.
cc @MaxGhenis @vahid-ahmadi
🤖 Generated with Claude Code
Facts request from the microcosm side (microcosm#840 assessment; register items microcosm#736 items 9/13; deferral tracker microcosm#796). Follow-on to #187/#188 (the April–December 2025 family-type and number-of-children histories).
Why
The four UC with-children calibration cells (
dwp.uc.households_single_with_children,…children_{1,2,5_or_more}) bind the Stat-Xplore Family Type and Number of Children splits, whose "with children" concept is not the one the model measures. Per the Stat-Xplore metadata:Measured on the current microcosm candidate (v18), UC households "with children" are 2.37m on the model's own-children concept but 3.06m on DWP's any-verified-under-20 concept, against DWP's 3.13m — the deficit the deferrals record is mostly a concept gap. The field that carries DWP's count on the concept the model computes (a child element in the award) is Child Entitlement (Yes/No): "whether a household receives a child element in their Universal Credit award … responsibility for a child or qualifying young person" (metadata).
What
Two new record sets from Stat-Xplore
UC_Households(count measureV_F_UC_HOUSEHOLDS), Great Britain, monthly April–December 2025 (same months as #188's packages so the calendar-average operation is sum-consistent):Optional third, if cheap in the same query: Family Type × Monthly Award Amount band "No payment" only, monthly — the nil-award households the model's
universal_credit > 0measure can never match (May 2025: 66,982 single-with-children, 73,849 couple-with-children, 501,957 in total; today only the May 2025 band package carries it — see the companion issue on band histories).Shape
Follow #188's pattern: facts-only, publisher values untransformed,
periodtyped monthly, benefit-unit entity, geography K03000001,assertion: observation, provenance to the archived API response (query echoed). Distinct source concepts or a child-entitlement dimension on the existingdwp.uc_householdsconcept — author's call under the loader guarantees; the consumer register will re-author its family-type and children references to whichever shape lands.Consumer
microcosm
uk_population_targets.json: the four family-type and five number-of-children cells move to "with child entitlement" counts (calendar-year average), which the engine measures directly (uc_child_element > 0/ thenum_uc_childrenvariable); the nil-award band becomes a value declaration on the same cells.cc @MaxGhenis @vahid-ahmadi
🤖 Generated with Claude Code