Body
Country: SE (Sweden)
Project root / allowed write root: /home/synthestat
Parent manager task: t_0c611b3b
Inputs to read first:
- /home/synthestat/workspace/manager_handoffs/SE_other_synthesis_ingest.md
- /home/synthestat/workspace/manager_handoffs/modeller/2026-05-19_1803_missing_requirements.md
- reviewer handoff on t_33ff07f7 (verdict NEEDS_MORE_SOURCES)
- docs/specs/research_knowledge_base.md
- docs/contracts/population_review_bundle.md
Mission:
Resolve the P0/P1 Sweden marginal/source gaps needed before a non-toy SE population review bundle. Focus on official/provenanced marginal target families, not implementation or modelling.
Scope / source families:
1. SCB DeSO/municipality/county/national targets referenced by other_synthesis: population_age_sex_deso, marital status, background, citizenship, birth region, municipality birth-country 2024, education 25-65, labour-market status, persons by building type and tenure, HushallDesoTyp / household type DeSO, households_by_type_children, households_by_size, persons_by_household_type_children_age, income quartiles, OD commuters, passenger cars.
2. Sweden geography assets for DeSO/RegSO/municipality/county mapping with boundary/reference periods.
3. Residential building/dwelling/home anchor source options: Lantmäteriet/SCB building/dwelling, HERE/local scaffold, OSM fallback. Mark official/licensed/proxy/scaffold/synthetic/unavailable.
4. Hidden-population marginal evidence gaps: homelessness, asylum/refugee/Ukrainian/Syrian/undocumented, institutional populations, seasonal workers, student housing/dormitories, dual residence beyond shared custody.
5. School/workplace/second-home marginal/source candidates only insofar as they affect population-bundle assignment scope.
Deliverable:
Write a concise source-research handoff under /home/synthestat/workspace/manager_handoffs/researcher/ named with timestamp and SE_population_marginals_sources. Include machine-readable enough detail for downloader: source IDs/table IDs, URLs/API endpoints, reference periods, geography levels, variables/classifications, retrieval constraints/licensing notes, quality flags, and whether each target is HARD/FIRM/SOFT/GUIDE/INFORMATIONAL candidate.
Do not overclaim. If a source is not found or is licensed/unavailable, say so explicitly. Hidden-population overlays must remain separate and uncertainty-aware; weak evidence means wider uncertainty or unavailable, not fake precision.
Definition of done:
- P0 source list is triaged into downloadable now / needs credentials / unavailable / needs human decision.
- Downloader-ready manifest recommendations are included.
- Concrete gaps that still block the modeler are called out.