Repository navigation
UK economy runs fail for every constituency, local authority and Northern Ireland #858
Description
Problem
UK economy runs fail for every constituency, every local authority and Northern Ireland. UK-wide, England, Scotland and Wales work.
GET https://api.policyengine.org/uk/economy/85546/over/1?region=<region>&time_period=2026 (reform 85546 = personal allowance £15,000; baseline 1 = current law) goes from computing to:
status: error
Simulation entrypoint execution failed: Simulation failed (correlation_id=…)
| Region | Result | correlation_id |
|---|---|---|
constituency/Aldershot |
error | df65c1f709ba4940a331e879287da0aa |
constituency/Sheffield Central |
error | 1eb1afc87df7… |
local_authority/Manchester |
error | 36479a66c0d04d8dbf905511bca173b3 |
local_authority/Hartlepool |
error | 60902c792bdd… |
country/ni |
error | 4515124b93a44395869ac05d6737df2e |
uk, country/england, country/scotland, country/wales |
ok |
Times: around 2026-10-20 18:35 UTC (the API logs the failure under policyengine-api; the cause is in the simulation worker).
Impact
Every local-level UK report and every Northern Ireland report fails, in app v2 (production, UK) and app v3.
Possibly related
The UK national result (policyengine-simulation-py6-5-0, model 2.130.1, data policyengine-uk-data 1.58.1) also has values that look wrong:
deep_poverty.all≥poverty.all(UK 14.9% vs 16.4%; Scotland 17.0% vs 16.1%; Wales 20.2% vs 19.3%), so the deep poverty measure looks wrong.- Wales: income decile 1 is empty (all
intra_decileshares and the average are 0). decile.relativeandwealth_decile.relativeare ×100 (Economy output: decile.relative is now a percentage (×100), not a fraction #827).
Activity
PavelMakarchuk commentedon Oct 20, 2026
Another UK local-area output problem, in a nation run that succeeds.
Report: app report 4684, region
country/wales, 2026, reformgov.hmrc.income_tax.rates.uk[0].rate0.20 → 0.21.constituency_impactfor Cardiff West:average_household_income_change: about −192relative_household_income_change: about +13.72 (the app shows +1372.3%)A tax rise gives a loss of £192, but the relative change is a large gain. The sign and the size do not agree. Both imply a baseline mean household income of about −£14 for Cardiff West, which is not possible.
Other Welsh constituencies are near −0.1% to −2.3%. Caerphilly is also high at −5.6%, so check it as well.
The app passes these values through without change (
societyWideOutputNormalizer.ts), so the cause is in the sim-api output. Possible cause: the constituency baseline income is computed on a small or wrongly weighted subset in a nation-filtered run.🤖 Generated with Claude Code
PavelMakarchuk commentedon Oct 23, 2026
Fixed by #881 (with policyengine-api#4024), now live. Checked on production after the 4024 deploy (23 Oct, 20:12 UTC; policyengine-uk 2.132.6, .py 6.10.0): report 4729 (UC freeze, 2027) runs UK-wide, and
breakdown=local_areasreturns 579 of 650 constituencies and 305 of 361 local authorities (the rest withheld by disclosure control), including all 18 Northern Ireland constituencies and 11 NI local authorities. National runs no longer build the local-area pair.