Choose the source before touching the estimator. A convenient variable in the wrong survey is still the wrong variable.
- Match the concept and its official definition.
- Match the record unit, universe, denominator, and exclusions.
- Check geography, period, and known series breaks.
- Confirm weights, replicate or design variables, imputations, and disclosure treatments.
- Confirm access and redistribution terms.
- Require a topic-specific official benchmark before publication.
not available, not found, not comparable, not implemented, and
credential required are different results. Do not collapse them into
"no data."
| Need | Strong candidates | Important boundary |
|---|---|---|
| Wealth and balance sheets | SCF | Family unit, five implicates, 999 bootstrap replicates |
| Housing stock and costs | AHS | Housing-unit design; modules and metro samples rotate |
| Spending | Consumer Expenditure | Interview and Diary instruments cover different concepts |
| Income, poverty, programs, labor, time use | ACS PUMS, CPS ASEC, SIPP, ATUS | Person, household, panel, and person-day records are not interchangeable |
| Health spending and coverage | MEPS, OECD SHA | Person-level expenditure is not a national-accounts financing measure |
| Financial experience and attitudes | SHED, GSS | Self-reports are not balance sheets; GSS output has separate terms limits |
| International macro comparisons | OECD, World Bank, Eurostat | Definitions, units, and revision policies differ by dataset |
| Electricity and energy | EIA Form 861, EIA/FRED energy watch | Utility census, market series, and pipeline capacity answer different questions |
uv run microdata sources
uv run microdata statusThe first command reads config/sources.yaml; the second reads the current
host's validated release store. Source-selection notes record definitions and
decisions, not a snapshot of one maintainer's machine.
Detailed notes live in docs/source-selection/.