General government spending as share of GDP, excluding transfers already captured under fiscal.transfer_expansion to avoid double-counting.
Size of cash and near-cash transfer programmes (unemployment benefits, means-tested assistance, universal child benefits). Architecturally distinct from forced-saving schemes — see condition welfare_architecture.
Tighter operating allowance than 2019-2022 Wellbeing Budgets, framed around cost-of-living support (prescription charges removed, 20 hours free ECE extended to two-year-olds, public transport free under-13s and half-price under-25s) and cyclone recovery spending after Cyclone Gabrielle. Last Labour budget before the October 2023 election loss.
Per invariant 3, reforms are scored by what they did on each channel-separated axis, not by the party that enacted them. This fingerprint is how the policy-match engine finds historical analogues.
Explicit links are curated by the author. Inferred links are hypotheses in the library that test the same axes this policy moved — the framework's answer to "what does the data say about a policy like this?".
Ranked by axis-fingerprint overlap with this policy. Direction match bolded — those are the closest historical analogues. Shape of the match is what drives policy-outcome comparison, not the country or party label.