Add child maintenance planning docs page (#669)

changelog.d/669.md

@@ -0,0 +1 @@
+- Add a planning docs page for child maintenance modelling under `docs/book/programs/gov/cms/child-maintenance.md`, flagging that the current generic `maintenance_income`/`maintenance_expenses` inputs miss the Universal Credit asymmetry highlighted by the CSJ "Hidden Parent Poverty Trap" paper, and outlining a three-phase implementation route (disambiguate inputs, fix the UC asymmetry, model the CMS formula).

docs/book/programs/gov/cms/child-maintenance.md

@@ -0,0 +1,133 @@
+# Child maintenance (planned)
+
+```{warning}
+**Not yet modelled in detail.** PolicyEngine UK currently has two generic
+input variables — `maintenance_income` and `maintenance_expenses` — that
+flow through HBAI but do **not** distinguish child maintenance from
+other maintenance and do **not** apply the Universal Credit asymmetry
+described below. This page captures the agreed scope for adding a
+properly-modelled Child Maintenance Service (CMS) calculation, tracked
+under [#669](https://github.com/PolicyEngine/policyengine-uk/issues/669).
+```
+
+## Why model child maintenance separately
+
+The Centre for Social Justice paper [*The Hidden Parent Poverty Trap:
+Child Maintenance and Universal Credit*][csj] shows that the
+**asymmetric** UC treatment of child maintenance produces a meaningful
+work-incentive cliff for paying parents:
+
+- **Receiving parent**: child maintenance income is **fully disregarded**
+  in the UC means test. That's good for the receiving household and
+  matches the policy intent of CMS.
+- **Paying parent**: child maintenance paid out is **not** deducted from
+  earnings in the UC means test. The paying parent is means-tested on
+  gross earnings as if the maintenance had never left their pocket.
+
+The current PolicyEngine treatment uses a single `market_income`
+deduction `income - maintenance_expenses` (`variables/household/income/
+market_income.py`). That is fine for HBAI net income but is **too
+generous** to the paying parent in UC — they would currently see
+maintenance reduce their UC-applicable income too, producing higher UC
+than they actually get.
+
+## Scope
+
+### Phase 1 — disambiguate the inputs
+
+- New input variable `child_maintenance_received` (Person, GBP, annual)
+  — the amount of CMS / private child maintenance the person receives.
+- New input variable `child_maintenance_paid` (Person, GBP, annual) —
+  the amount the person pays out.
+- Keep `maintenance_income` / `maintenance_expenses` for spousal and
+  other maintenance flows; document the distinction explicitly.
+- In `policyengine-uk-data`, route the FRS child-maintenance items into
+  the new variables and the FRS spousal-maintenance items into the
+  existing ones.
+
+### Phase 2 — fix the UC asymmetry
+
+- New variable `uc_applicable_child_maintenance` (BenUnit) that returns
+  **`child_maintenance_received` only**, with `child_maintenance_paid`
+  set to **zero** — i.e. the UC-side asymmetry.
+- Subtract `uc_applicable_child_maintenance` from
+  `uc_applicable_income` so it's fully disregarded for the receiver.
+- Crucially **do not** subtract `child_maintenance_paid` from
+  `uc_applicable_income`; this reproduces the CSJ-identified poverty
+  trap.
+
+### Phase 3 — model the CMS formula itself
+
+Currently both `maintenance_income` and the proposed `child_maintenance_*`
+variables are pure inputs. Phase 3 would derive the CMS award when the
+required inputs are present, using the gross-income basis the CMS
+applies:
+
+| Paying-parent gross weekly income | Rule |
+|------------------------------------|------|
+| Less than £7  | No payment ("nil rate") |
+| £7 – £100     | Flat-rate £7 |
+| £100 – £200   | Reduced rate (formula-based) |
+| £200 – £3,000 | Basic rate: 12% / 16% / 19% of gross income for 1 / 2 / 3+ children, with a tapered top slice between £800 and £3,000 |
+| Over £3,000   | Capped at the £3,000 figure (statutory upper limit) |
+
+Plus the **shared-care reduction** (one-seventh per night of shared
+care up to 174 nights/year) and the **other-children adjustment** for
+non-qualifying children living with the paying parent.
+
+Phase 3 is optional for headline UC analysis (where the asymmetry in
+Phase 2 is what matters) but unlocks reform scenarios that change the
+CMS rate or threshold.
+
+## Implementation outline
+
+### New parameters under `gov/cms/`
+
+- `gross_income_band/{lower,upper}_threshold.yaml` (£7 and £3,000 weekly
+  in 2025 terms)
+- `rate/basic/{1,2,3_or_more}_child.yaml` (12 / 16 / 19%)
+- `rate/reduced/{base,marginal}.yaml`
+- `shared_care/reduction_per_night.yaml`
+- `take_up/cms_collect_share.yaml` (HMRC published share of CMS-vs-
+  family-arrangement maintenance)
+
+### New variables under `variables/gov/cms/`
+
+- `child_maintenance_gross_weekly_income` (Person)
+- `child_maintenance_obligation_pre_shared_care` (Person)
+- `child_maintenance_shared_care_reduction` (Person)
+- `child_maintenance_paid_cms_basis` (Person, derived award)
+
+### UC integration
+
+- `variables/gov/dwp/universal_credit/uc_applicable_income.py` adds
+  `child_maintenance_received` as a deduction (full disregard) and
+  **does not** deduct `child_maintenance_paid` — matching the policy.
+- HBAI keeps the current symmetric treatment; the CSJ asymmetry is a
+  UC-side specific.
+
+## Data needs
+
+- **FRS** records both receipt and payment of maintenance and identifies
+  the recipient/payer at person level. Phase 1 imputation routes FRS
+  child-maintenance items into the new variables.
+- **CMS / Child Maintenance Statistics** ([gov.uk/government/collections/
+  child-maintenance-service-statistics][cms-stats]) give published
+  CMS-arrangement caseload and total maintenance flowing through the
+  statutory scheme; useful for calibrating the new variables against
+  the share of maintenance that goes through CMS vs. private
+  ("family-based") arrangements.
+
+## References
+
+- Centre for Social Justice, [The Hidden Parent Poverty Trap: Child
+  Maintenance and Universal Credit][csj].
+- DWP, [How we work out child maintenance][how-we-work-out] — official
+  user-facing CMS methodology guide.
+- [Child Support Act 1991](https://www.legislation.gov.uk/ukpga/1991/48/contents) — primary statute.
+- [The Child Support Maintenance Calculation Regulations 2012 (SI 2012/2677)](https://www.legislation.gov.uk/uksi/2012/2677/contents) — the operational regulations.
+- HMRC/DWP, [Child Maintenance Service statistics][cms-stats] — caseload and expenditure outturns.
+
+[csj]: https://www.centreforsocialjustice.org.uk/library/the-hidden-parent-poverty-trap-child-maintenance-and-universal-credit
+[how-we-work-out]: https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/672432/how-we-work-out-child-maintenance.pdf
+[cms-stats]: https://www.gov.uk/government/collections/child-maintenance-service-statistics