Hold the Child Tax Credit family element at £545 and apply the 6 April 2017 birth-date condition

[MaxGhenis]

Summary

The Child Tax Credit family element was £545 a year for the whole life of tax credits, 6 April 2003 to 5 April 2025. PE-UK started the parameter in 2016 and CPI-uprated it, so it read £550.99 in 2017-18, rising to £721.57 in 2024-25. The model also paid it to every eligible family, though from 6 April 2017 it is only for families with a child or qualifying young person born before 6 April 2017.

This PR:

Change	Law
elements/family_element.yaml: £545 from 2003-04-06, uprating removed	CTC Regs 2002 reg 7(3)(b) as made; reg 7(3) as substituted by SI 2011/1035 reg 2(2)
New elements/family_element_baby_addition.yaml: £545 more where any child is under one, 2003-04-06 to 2011-04-05	reg 7(3)(a) as made (£1,090)
New eligibility/family_element_born_before.yaml: 20170406	TCA 2002 s.9(2)(a) (WRWA 2016 s.13(2)(a)); CTC Regs reg 7(2)(a) (SI 2017/387 reg 4(b))
CTC_family_element applies the cutoff and the pre-2011 baby rate	as above

The law

Fetched verbatim from legislation.gov.uk /data.xml on 2026-10-02.

CTC Regs 2002 reg 7(3), as made (in force for awards from 6 April 2003, reg 1):

The family element of child tax credit— (a) in a case where any of the children referred to in paragraph (2)(b) above is under the age of one year, is £1,090; and (b) in any other case, is £545.

SI 2011/1035 reg 2(2) (Tax Credits Up-rating Regulations 2011, awards for the year from 6 April 2011):

For paragraph (3) substitute— (3) The family element of child tax credit is £545.

Never uprated.

• Apart from that 2011 substitution, the Tax Credits Up-rating Regulations from 2004/941 to 2024/247 amend only reg 7(4) (individual element) and (5) (disability element).
• legislation.gov.uk's changes feed for SI 2002/2007 records 107 effects on reg 7, and the substitution is the only one on paragraph (3). The feed is not a complete list: it omits SI 2008/796 reg 2, which amends paragraph (4) only (fetched separately).
• Point-in-time texts of reg 7(3) for 2003-04, 2010-11, 2011-12, 2016-17, 2017-18 and 2024-25 all agree.
• The 2016-20 tax credit freeze (WRWA 2016 s.12 and Sch 1 para 2(b)) covers reg 7(4)(c) and (f), not paragraph (3).
• HMRC's rates page gives £545 for 2023-24 and 2024-25, and "not applicable" for 2025-26.

Reg 7(2)(a) from 6 April 2017 (words inserted by SI 2017/387 reg 4(b)):

the family element of child tax credit if the claimant is, or either or both the joint claimants are, responsible for a child or qualifying young person who was born before 6th April 2017, and

TCA 2002 s.9(2)(a), as substituted by WRWA 2016 s.13(2)(a) (in force 6 April 2017, s.36(4)), says the same. SI 2017/387 has no transitional saving for existing awards.

Two corrections to the brief that led to this PR: the rule is in reg 7(2)(a), and reg 7(2A) is the child limit. And the amount was £545 from 2003; what changed in 2011 was the end of the £1,090 rate for a family with a baby.

Model notes

• Birth dates. On main a person's date of birth is known only to the year (birth_year, the period less age, or an input). A child counts as born before the cutoff when 6 April of their birth year is before 20170406. That means a birth year of 2016 or earlier, exactly as is_CTC_child_limit_exempt reads the 2017 child limit. Apply the 6 April 2017 child cutoffs by date of birth #1904 (date of birth from age and months since the last birthday) will replace this with its birth_day helper; the parameter is already a YYYYMMDD date to match its born_before parameters.
• Partial birth-year inputs. A situation that enters birth_year for some people leaves everyone else at 0 (core fills the input array). Read naively, 0 is "born before 2017". The formula treats 0 as not given and uses the period less age. The Hypothesis characterisation found this case, and it has a YAML case. is_CTC_child_limit_exempt, uc_is_child_born_before_child_limit and the Pension Credit child addition share the issue on main. Apply the 6 April 2017 child cutoffs by date of birth #1904 fixes those by moving them to its helper, which ignores birth_year inputs and treats a 0 date_of_birth as not given.
• Under one. The baby rate reads age < 1 for the year. In law it ran by relevant period (daily), so in a whole-year model a child who turns one mid-year keeps it all year. It affects only 2003-04 to 2010-11.
• Years before 2015. PE-UK converts parameters to fiscal years for 2015 to 2040 only. An earlier YEAR period reads parameters at 1 January, which is in the previous tax year. The pre-2015 YAML cases use periods 2004, 2010 and 2012, where both tax years have the same law. The individual element and income thresholds start in April 2016 in the parameter files, and the model backdates each parameter's earliest value to 1 January 2015, so neither has values before 2015. Follow-up chip: extend the conversion.

Invariants (stated and tested)

policyengine_uk/tests/test_ctc_family_element_properties.py:

1. Parameter, exhaustive. On every day from 2003-04-06 to 2040-12-31 the family element is £545. With the baby addition it is £1,090 before 2011-04-06 and £545 from then. In every converted fiscal year (2015 to 2040) the YEAR value is £545, the baby addition is 0 and the cutoff is 20170406.
2. One flat amount, for the right families (Hypothesis, 2015-16 to 2024-25). The populations mix 1-2 adults, 0-4 dependants aged 0-21 (with or without non-advanced education), entered or missing birth years, and CTC eligible or not. The family element is always 0 or £545, never per child. It is £545 exactly when the family is CTC-eligible and counts a child or qualifying young person born in 2016 or earlier. The oracle computes birth years from the generated data, not from the model.
3. Order invariance (Hypothesis): reversing each family's member order changes nothing.
4. Metamorphic (Hypothesis, 2017-18 on): adding a newborn never changes the family element.
5. Baby rate (Hypothesis, 2004 to 2010): £1,090 when an eligible family counts a child under one, otherwise £545.

Tests

policyengine_uk/tests/policy/baseline/gov/dwp/tax_credits/ctc_family_element.yaml: 22 cases, worked from the law in the file header. On main, 20 fail on value or the missing parameters. The two that pass are 2016-17 (where main already had £545) and the 2025-26 zero award. For example, in 2024-25 a lone parent on £25,000 with one child born in 2014 gets £3,455 + £545 - 41% x (25,000 - 19,995) = £1,947.95 (main: £2,124.52). With one child born in 2019 the award is £1,402.95. Awards use the model's annual arithmetic. SI 2002/2008 reg 8(3) steps 1-2 take each element as a daily rate rounded up to the penny, so a full-year statutory award is a few pounds higher (£1,952.00 for the first case). The element amounts and the condition are exact.

test_ctc_family_element_properties.py: 31 passed locally. The characterisation property pins a partial birth-year example in both directions. The full YAML and pytest suites run in CI's Test job.

Enhanced FRS impact (2024-25)

One real Microsimulation per package state on a copy of enhanced_frs_2024_25.h5 (sha256 e433e532…): base origin/main 15de0cadf, branch b34cc9478. Scripts and aggregate outputs are kept with the session notes. Weighted totals, 2024-25:

	Main	This PR	Change
Family element in CTC maximum rates	£15.2m	£9.1m	−£6.1m
Child Tax Credit paid	£80.8m	£78.6m	−£2.19m
Housing Benefit (tax credits are HB income)			+£0.19m
Household net income			−£2.00m
Poverty rates (BHC, AHC, children, pensioners)			unchanged

Where the family element moves:

• 21,000 weighted families are CTC-eligible in the data.
  • 16,600 keep it, at £545 instead of £721.57: −£2.94m in maximum rates.
  • 4,400 lose it because every child was born on or after 6 April 2017: −£3.17m.
  • None gain it.
• CTC paid falls by less (−£1.77m and −£0.42m). Families still paid CTC on both bases take the full element change (−£2.19m). The other −£3.92m of element falls on families tapered to zero on both bases.
• No family stops getting CTC altogether.

The amounts are small because the data has few tax credit awards left in 2024-25: 21,000 weighted CTC-eligible families and £80.8m of CTC paid on main. In 2025-26 tax credits pay nothing on either base. Maximum rates and pre-minimum CTC still change, but abolition stops those changes reaching paid CTC, Housing Benefit or household net income.

Coordination

• Set DWP benefit rates to the announced amounts, not CPI projections #1925 (DWP rates to announced amounts) also removes uprating from family_element.yaml and adds a comment. This PR supersedes that hunk: it keeps Set DWP benefit rates to the announced amounts, not CPI projections #1925's reference entries and adds the law. Whichever merges second should take this PR's version of the file.
• Apply the 6 April 2017 child cutoffs by date of birth #1904 (6 April 2017 cutoffs by date of birth) should move CTC_family_element onto its birth_day helper (birth_day(person, period) < ctc.eligibility.family_element_born_before). Its whole-age invariant (born before 6 April 2017 iff period - age < 2017) then covers this reader too.
• WTC 30-hour element is CPI-uprated instead of using the 2024-25 rate #1759 (WTC 30-hour element uprating) is fixed by Set DWP benefit rates to the announced amounts, not CPI projections #1925's worker.yaml values, not here.

axiom: TheAxiomFoundation/rulespec-uk#372 queued (extended with reg 7(3) as made and as substituted from 2011, plus companion tests)

Fixes #2050

🤖 Generated with Claude Code

[MaxGhenis]

#1904 merged into main as e24c8668f. It changes two things this PR relies on.

birth_year

• birth_year is now the calendar year of an exact date of birth: utils.dates.birth_day(person, period), from age and months_since_last_birthday, or a date_of_birth input.
• For whole ages in household situations, it is still period - age.
• In survey data, people with birthdays from 7 October to 31 December now get the year before.

Exact comparison for the family element

• The s.9(2)(a) family element condition can now compare dates exactly. Apply the 6 April 2017 child cutoffs by date of birth #1904's CTC child limit does it this way:

  from policyengine_uk.utils.dates import birth_day
  born_before = birth_day(person, period) < limit.born_before

  Here limit.born_before is gov.dwp.tax_credits.child_tax_credit.limit.born_before, 20170406.

• That avoids a year proxy, which treats the whole cohort born from October 2016 to October 2017 as born after the cutoff.

• To set an exact date in YAML tests, use date_of_birth: 20170405 with age as the legal age on 6 October.

• birth_year inputs no longer drive any cutoff.

• A date_of_birth input for only some people leaves the others placed by age. Inputs that are not dates, or that contradict age, raise an error.

Your branch

• This PR's family-element formula and tests still work as written after merging main. Its YAML and property tests set birth_year inputs, and a birth_year input still overrides birth_year.
• Switching the condition to birth_day would mean moving those tests to date_of_birth inputs.
• This PR doesn't read the removed limit.start_year, so that removal doesn't touch it.