Lanes: packages under ledger-source-ingestor; the sa2 geography level under ledger-contract-maintainer
Depends on: conventions from #175/#176; not started until epic microcosm#343 opens local-area scope (v0 doctrine: national first; local areas filter one national dataset)
Context
Filed for scope visibility, mirroring UK wave 3 (#134). NZ local-area calibration would back TA/SA2-grain targets. Unlike the UK there is no local target profile declared yet — profile extension follows the packages.
Candidate packages
| Backing |
Package(s) |
Source mechanics |
| Census 2023 totals by topic at SA2 |
stats_nz/census_2023_individuals_sa2, stats_nz/census_2023_households_sa2 |
ArcGIS hub CSV part-files; ~2,300 SA2s — whole-table orientation mandatory, rows: geometry only |
| MBIE quarterly rents at SA2 |
mbie/tenancy_bond_rents_sa2 |
detailed-quarterly-tenancy-2020-to-2026.csv (11 MB, SA2 grain), selected_rows restriction |
| LEED earnings by TA |
stats_nz/leed_earnings_by_ta |
ADE curated extract (mean/median earnings, filled jobs by TA, quarterly) |
| HUD dashboard TLA cuts |
extend hud/housing_dashboard_* (#177) |
Social-housing register + AS by TLA |
| Subnational projections (2023-base) |
stats_nz/subnational_projections_2023_base |
Only if populace targets forward local years; assertion: source_projection |
Contract dependency
A first-class sa2 geography level (deliberately not statistical_scope — SA2 is a general statistical geography like msoa), with vintage sa2_2023. One small contract-lane PR, the msoa precedent.
Acceptance
Every declared local family resolves from a consumer artifact at its geography level with full lineage; the first SA2 package settles the row-spec form for ~2,300-row tables and documents it in docs/agent-source-package-harness.md if not already covered; coverage vs the (future) local profile enumerated with signed exclusions — no silent drops.
Sizing
~5–6 packages ≈ 2–3 PRs + 1 contract PR. The SA2 census packages are the expensive ones — budget as UK wave 3 did.
Lanes: packages under
ledger-source-ingestor; thesa2geography level underledger-contract-maintainerDepends on: conventions from #175/#176; not started until epic microcosm#343 opens local-area scope (v0 doctrine: national first; local areas filter one national dataset)
Context
Filed for scope visibility, mirroring UK wave 3 (#134). NZ local-area calibration would back TA/SA2-grain targets. Unlike the UK there is no local target profile declared yet — profile extension follows the packages.
Candidate packages
stats_nz/census_2023_individuals_sa2,stats_nz/census_2023_households_sa2rows:geometry onlymbie/tenancy_bond_rents_sa2detailed-quarterly-tenancy-2020-to-2026.csv(11 MB, SA2 grain),selected_rowsrestrictionstats_nz/leed_earnings_by_tahud/housing_dashboard_*(#177)stats_nz/subnational_projections_2023_baseassertion: source_projectionContract dependency
A first-class
sa2geography level (deliberately notstatistical_scope— SA2 is a general statistical geography likemsoa), with vintagesa2_2023. One small contract-lane PR, themsoaprecedent.Acceptance
Every declared local family resolves from a consumer artifact at its geography level with full lineage; the first SA2 package settles the row-spec form for ~2,300-row tables and documents it in
docs/agent-source-package-harness.mdif not already covered; coverage vs the (future) local profile enumerated with signed exclusions — no silent drops.Sizing
~5–6 packages ≈ 2–3 PRs + 1 contract PR. The SA2 census packages are the expensive ones — budget as UK wave 3 did.