Fix state pension reforms having no budget impact (#1178)#1509
Fix state pension reforms having no budget impact (#1178)#1509
Conversation
basic_state_pension and additional_state_pension read parameters at data_year using node-level access (parameters(instant).path.amount), which doesn't reflect parameter reforms. Fixed by: 1. Using leaf-level access (parameters.path.amount(instant)) which correctly reflects reforms applied to the parameter tree. 2. Computing each person's share of the maximum at data_year, then applying that share to the (possibly reformed) maximum at the simulation period — replacing the manual triple-lock uprating. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
vahid-ahmadi
left a comment
vahid-ahmadi
left a comment
There was a problem hiding this comment.
1. new_state_pension has the same bug
new_state_pension.py:18 uses node-level access (parameters(period).gov.dwp...), which I verified empirically does not see reforms — same root cause fixed here. Currently dormant (£0 in baseline) but trivial to fix with the same leaf-level pattern.
2. additional_state_pension baseline changes should be documented
The old formula had no uprating; the new one adds bsp(period) / bsp(data_year) (~1.26x at 2029). This increases baseline additional SP by ~26% (~£5bn) and is what shifts the UC taper expected impact from -34.3 to -42.0. Adding uprating is arguably correct (real additional SP is CPI-uprated), but the PR description says the ratio is 1.0 when no reform is active — that only holds when data_year == period. Worth acknowledging as an intentional improvement.
|
Superseded by #1548, which reapplied this fix on current main, passed fresh CI, and was merged. Closing the stale conflicting branch. |
2 participants
Summary
basic_state_pensionandadditional_state_pensionformulas now respond to parameter changes at the simulation periodRoot cause
basic_state_pensionreadparameters(data_year)instead ofparameters(period). The formula fetched the parameter at the survey data year and applied a fixed triple-lock uprating, so reforms to the parameter for the simulation year were never seen. Same issue inadditional_state_pension.Fix
Both formulas now compute a reform ratio:
parameters(period).amount / parameters(data_year).amountapplied as a multiplier on top of the triple-lock-uprated amount. If no reform is active, the ratio is 1.0 (no change). If a reform modifies the parameter, the ratio adjusts the output accordingly.Test plan
Fixes #1178