Enhanced FRS 2024-25 State Pension is £11.6bn below the FRS's own admin-linked estimate #493

State Pension in the enhanced FRS 2024-25 is well below both the FRS's own grossed estimate and DWP's administrative figures, in the data year itself. Found while fixing PolicyEngine/policyengine-uk#1921 (PolicyEngine/policyengine-uk#1922); part of PolicyEngine/policyengine-uk#1632.

Figures, 2024-25

The model row is the weighted sum of state_pension_reported in the data year (equal to state_pension in 2024-25) and the weighted count of people with state_pension > 0.

Sources:

• FRS 2024-25 methodology and standard error tables, M.6a row 19 and M.6b row 15: https://assets.publishing.service.gov.uk/media/69c3c0d2471d520038d0f571/ch1_methodology_and_standard_errors.xlsx
• FRS "integrating administrative data for benefits" tables, sheet 16. SP, FYE2025: https://assets.publishing.service.gov.uk/media/69c416acb66ff902f45441f0/family-resources-survey-transformation-benefits.xlsx
• From FYE2025, the FRS replaces benefit amounts with linked administrative records for linked respondents.

So the FRS itself gets within about 3% of DWP. The enhanced FRS is £11.6bn below the FRS's own UK estimate, with both fewer recipients and a lower mean amount.

Where to look

• datasets/frs.py (origin/main b45c373, lines 1151-1182) sets state_pension_reported to the BENEFITS table's benamt for benefit code 5, times WEEKS_IN_YEAR. Check whether that is the admin-linked amount in the 2024-25 FRS, and whether other State Pension-type codes (e.g. Graduated Retirement Benefit, or a separate additional pension code) are missed.
• On the Survey of Personal Incomes donor rows (the zero-weight copy that calibration can upweight), state_pension_reported is overwritten by a second-stage QRF: it is in FRS_ONLY_PERSON_VARIABLES (datasets/imputations/frs_only.py:85, tag 1.57.4), applied by impute_frs_only_variables from imputations/income.py. Compare the recipient count and mean amount on those rows with the FRS rows after calibration.
• Compare weighted recipient counts and mean amounts before and after calibration, to split the £11.6bn into weights and amounts.
• Check which State Pension calibration target is used, if any (storage/tax_benefit.csv has an OBR March 2024 row).