bsp_reported is CPI-uprated but BSP rates are frozen, and the variable is mostly War Widow's Pension #2381

Summary

bsp_reported is projected forward with CPI, but Bereavement Support Payment (BSP) rates have been frozen since BSP started on 6 April 2017. The catch is that bsp_reported is not only BSP. Both data pipelines build it from FRS benefit codes 6 and 9, and code 9 is War Widow's/Widower's Pension, which is uprated by CPI every April. In the FRS 2024-25 microdata, code 9 is about 90% of the value of bsp_reported. So simply freezing the variable would make the projection worse, not better.

The fix is to split the war widow's pension out into its own CPI-uprated variable, and then freeze BSP once the datasets populate the split.

The law

• BSP rates. The Bereavement Support Payment Regulations 2017 (SI 2017/410), reg 3 set the higher rate at £3,500 plus £350 a month and the standard (lower) rate at £2,500 plus £100 a month. The revised text on legislation.gov.uk still carries those amounts.
• The up-rating order holds them flat. The Social Security Benefits Up-rating Order 2026 (SI 2026/148), art 19 says each of £350, £3,500, £100 and £2,500 "remains unchanged". The DWP Benefit and pension rates 2026 to 2027 table shows the same four amounts for both 2025/26 and 2026/27.
• Widowed Parent's Allowance (WPA) is uprated by CPI under art 18 of the same order ("Rates of bereavement benefits"): £150.90 a week in 2025/26 and £156.65 in 2026/27, a 3.8% rise. It is closed to deaths from 6 April 2017 onwards.
• War Widow's/Widower's Pension is paid under the War Pensions Scheme, not the social security up-rating order. It is uprated each April by The Naval, Military and Air Forces Etc. (Disablement and Death) Service Pensions (Amendment) Order in line with September CPI: 1.7% for 2025/26 (HB circular A2/2025) and 3.8% for 2026/27.

The model

• policyengine_uk/data/uprating_indices.yaml:42 lists bsp_reported under gov.economic_assumptions.yoy_growth.obr.consumer_price_index, which is what projects the dataset column forward.
• policyengine_uk/variables/gov/dwp/bsp_reported.py also declares uprating = "gov.economic_assumptions.indices.obr.consumer_price_index". policyengine-uk-data builds storage/uprating_factors.csv from that attribute, so its bsp_reported row is CPI too (1.0 in 2020 rising to 1.38 by 2029).
• There is no formula for BSP and there are no BSP rate parameters. bsp (policyengine_uk/variables/gov/dwp/bsp.py:11) is just adds = ["bsp_reported"].
• bsp feeds household_benefits (and so household_net_income), hbai_benefits, hbai_household_net_income, gov_spending, pre_budget_change_household_benefits and the disable_simulated_benefits reform.

What the FRS codes are

The FRS 2024-25 benefits data dictionary labels code 6 "Bereavement Support Payment / Widowed Parent's Allowance" and code 9 "War Widow's/Widower's Pension". Both pipelines sum the two:

• policyengine-uk-data policyengine_uk_data/datasets/frs.py:1464: benefits.benefit.isin((6, 9))
• microcosm packages/microcosm-build/src/microcosm/build/uk_runtime/frs_spine.py:825: benefit.isin((6, 9))

Grossed with gross4, the raw FRS 2024-25 benefits table gives:

So about £100m (8%) of the reported total is BSP at the frozen rates, and about £1,093m (91%) is WPA and war widow's pension, both of which are uprated by CPI. The 141,600 grossed code 9 recipients look high for the war widow's caseload, so part of code 9 may be misreported survivors' pensions, but those are also uprated. BSP lump sums are not captured in either code; that is out of scope here.

VAR2 on code 6 rows is meant to carry the Wid question (1 BSP, 2 WPA), but in the 2024-25 EUL file it does not line up with the amounts (most £23.01 rows have VAR2 = 2), so an amount-based or admin-based split may be safer than VAR2.

Size of the effect

The CPI index in the parameters rises by a factor of 1.1225 from 2024 to 2029 (OBR year-on-year growth 3.4%, 2.3%, 2.0%, 2.0%, 2.0% for 2025 to 2029). Every pound of BSP in a 2024 dataset is therefore shown as £1.12 in 2029-30, when it should still be £1.00.

On the enhanced FRS 2024-25 dataset, bsp_reported totals £731.4m in 2024 and £840.2m in 2029 (about £92m of that rise is the CPI projection; the rest is weight growth). Using the raw FRS shares above:

• Keep CPI on the combined variable (status quo): the BSP part, about 8%, is overstated by about £8m in 2029-30.
• Freeze the combined variable: everything else, about 92% and almost all war widow's pension and WPA, is understated by about £84m in 2029-30.

Freezing the whole variable now would swap an £8m error for one ten times larger.

Proposed fix

1. policyengine-uk (this repo). Add war_widows_pension_reported (CPI-uprated, listed in uprating_indices.yaml) and war_widows_pension, and add war_widows_pension to every aggregate that lists bsp. Document on bsp_reported that it currently holds FRS codes 6 and 9, and that it stays on CPI until the datasets split code 9 out. Existing datasets have no war_widows_pension_reported column, so results do not change.
2. policyengine-uk-data and microcosm. Write code 9 to war_widows_pension_reported and only code 6 to bsp_reported. Ideally also move WPA (the code 6 rows at the WPA rate) out of bsp_reported, for example into its own variable, so that bsp_reported is BSP only.
3. policyengine-uk, follow-up. Once released datasets carry the split, drop bsp_reported from the CPI list in uprating_indices.yaml and remove its uprating attribute, so BSP stays at the 2017 rates. Then regenerate uprating_factors.csv in policyengine-uk-data, which will drop the bsp_reported row.

Freezing in step 3 before step 2 lands would put the £84m error above into results, so the order matters.

🤖 Generated with Claude Code