PolicyEngine / PolicyEngine/policyengine-uk-data

Scottish/Welsh council tax ~20% above devolved budget figures; WAS owned_land imputation unstable across releases

Open
#448 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
1
Forks
2
Avg merge
1d 14h
Merged PRs (30d)
4

Description

Two data-quality issues surfaced while auditing the geographic results of the council-tax-to-LVT paper (PolicyEngine/uk-lvt-paper). Both materially distort nation-level distributional results.

1. Scottish and Welsh council tax overstated

On enhanced_frs_2023_24.h5@1.55.10 (2026, policyengine-uk 2.88.20), modelled council_tax aggregates by nation:

Nation Model gross Model net Official comparator
England £53.8bn £50.5bn ≈£45.6bn requirement 2025–26 (within a few % once uprated to 2026–27)
Scotland £4.8bn £4.0bn ≈£3.1–3.4bn
Wales £3.3bn £3.1bn ≈£2.4–2.6bn

The same pattern holds on enhanced_frs_2024_25.h5 (Scotland £4.3bn gross). Scottish and Welsh liabilities run roughly 20–30% above devolved budget figures while England is close. Average modelled net bill in Scotland (£1,458) is implausibly near England's (£1,872) given Scottish band D (~£1,421 in 2024–25, ex-water) and Scotland's band distribution being skewed to A–C. Possible causes worth checking: band-rate averages for Scotland including water/sewerage charges, or the VOA/SAA band calibration weighting.

Downstream effect: any reform crediting households with their abolished council tax (e.g. an LVT swap) overstates Scottish/Welsh gains by £200–300/household/year — enough to flip marginal constituencies from loss to gain.

2. owned_land (WAS direct land holdings) is large and unstable across releases

household_land_value = property_wealth × λ_r + owned_land. The owned_land component:

Release UK total Wales avg/hh Wales share of Welsh household land Scotland share
enhanced_frs_2023_24@1.55.10 £0.67tn £53,548 43.6% 27.6%
enhanced_frs_2024_25 (local) £0.46tn £11,680 9.8% 31.7%

A 4.6× swing in average Welsh direct landholding between adjacent releases points to a handful of large WAS farm/estate records being imputed into different donors each time. At £0.46–0.67tn this is 9–12% of household land — not noise. Suggest: winsorising the WAS land-value donor variable, imputing it separately from the main wealth block, or at least a regression test on nation-level owned_land aggregates.

Related: PolicyEngine/policyengine-uk#1808 (NI land intensity 0.673). Context: PolicyEngine/uk-lvt-paper#4.

🤖 Generated with Claude Code

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Reproduce the nation-level council_tax and owned_land aggregates from enhanced_frs_2023_24.h5@1.55.10 and enhanced_frs_2024_25, separating Scotland, Wales, England, and the Wales and Scotland shares. Trace the WAS direct landholding imputation and the council-tax band-rate or calibration inputs, then add regression coverage for the reported aggregates across releases; done means the discrepancies are explained and the relevant totals remain stable or are intentionally documented.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
data-engineering
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.