Adjudicate the incumbent's four post-calibration aggregate scalers #790

Parent: #665 (WS-D/E boundary). Extracted from the 2026-08-14 adjudication request on #623 so the ruling has a home that survives #623 closing at the first certified cut; registered in #736.

The question

The incumbent runs four aggregate adjustments after calibration, on weighted sums:

1. Rail-subsidy scaling to the ORR total
2. Bus-subsidy scaling to the DfT total
3. Fuel-litre-proxy scaling to HMRC road-fuel clearances
4. Bus-fare rescale to the DfT total, which also reshapes the income gradient to NTS trips-per-quintile — a distributional operation, not only an aggregate one

Because they read calibrated weights, they cannot live in the WS-E source stages — the ports (#682 consumption/services, #685 bus fares) deliberately exclude them — and the calibration seam landed without them (#743, per the 2026-08-20 scope ruling). Nothing in the migrated pipeline reproduces their effect today.

The options

• (a) Port them as a declared post-calibration step in the seam, under adjudicated doctrine (each scaler a declared operation with its target total pinned as a chronicle fact, mass semantics recorded). Keeps the incumbent's published levels; adds a second adjustment layer after the solve.
• (b) Replace them with calibration targets — rail/bus subsidy totals and road-fuel volumes as chronicle facts the solver binds directly, letting reweighting do what the scalers did by multiplication. No post-solve mutation; distributional NTS gradient would need its own representation (quintile-level facts) or a signed limitation.
• (c) Sign their absence — accept the level gap on these four surfaces with tracked exclusions until (a) or (b) lands.

Evidence that bears on the ruling (2026-08-26, #757 branch)

The obr.fuel_duties investigation on the first live seam runs measured the frame carrying ~37% of implied road-fuel litres (LCFS two-week-diary zero-inflation: fuel-recording share 0.44 petrol / 0.19 diesel against ~78% car ownership), leaving obr.fuel_duties at −62.5% relative error. The incumbent's fuel-litre scaler to HMRC clearances compensated for exactly this under-capture. The scaler ruling and the pending fuel-duties disposition (household-incidence retarget vs total-receipts fact, flagged on #757) are the same question from two sides: whether the gap closes by scaling values, by binding volume targets, or by re-scoping the fact to what a household frame can carry.

Related: rail levels measured ~3.07× donor at #686 (after accounting for the incumbent's per-head division defect, policyengine-uk-data#469), so the rail scaler's effect is entangled with a known incumbent defect on the same surface.

Acceptance

• A ruling choosing (a), (b), or (c) per scaler — they need not share one answer; the NTS gradient reshape is adjudicated separately from the three pure aggregate scalings.
• The chosen mechanism lands with declared provenance (facts pinned, mass/distributional semantics recorded) or signed exclusions with tracking.
• The obr.fuel_duties disposition on Swap the microcosm-built UK spine into the national build and retire the incumbent feed #757 is resolved consistently with the fuel-scaler ruling.

🤖 Filed with Claude Code