Source register

Akashic combines public election, Census, geographic, religion, registration, migration, and officeholder sources. This page is an orientation to the principal source families; the record-level provenance manifest remains the final source for a specific figure.

Presidential elections

Source family	Primary use	Important notes
ICPSR historical election archives	Early county presidential returns	Sparse in the earliest period; 1876–1888 remains partial
MIT Election Data and Science Lab	Modern county/state/national election series and candidate returns	Used without altering certified vote totals; party/candidate normalization is documented
State election authorities	Recent certified official results, including 2024	State-specific formats and certification dates vary
VEST precinct files	Precinct geometry and election returns	Coverage and field conventions differ by state/cycle; rollups are validated against certified totals
MEDSL precinct collections	Additional precinct and downballot coverage	Candidate and office schemas require normalization and reconciliation
OpenElections and state-specific files	Gap filling where licensed and verified	Used only when the source and reconciliation contract are documented

Akashic does not use newswire election-night totals as the permanent certified historical record.

Demographics and geography

Source family	Primary use	Important notes
US Census Bureau decennial products	Population and geographic foundations	Geography vintage must match the analysis
American Community Survey five-year products	Income, age, race/ethnicity, ancestry, language, education, poverty, and related measures	Each field has a table/universe and survey vintage; margins of error may matter even when not displayed
Census TIGER/Line and cartographic boundaries	Geographic identifiers and shapes	Boundary simplification may be used for display while analytical rollups use more detailed source geometry
Census geographic crosswalks	Parent relationships and current identifiers	Crosswalk method matters when historical results are presented on modern boundaries

Religion

The US Religion Census / ASARB source family supports religion-adherence context. Counts are tied to congregation location and may understate resident religious participation in small or sparse counties whose residents travel to a regional hub. Akashic may provide commuting-zone service-shed context; that context is not a direct county observation.

Voter registration

Registration data comes from state election agencies or official state reports. States differ in whether they register voters by party, whether inactive registrations are reported, and how often files are updated. A missing party split in a no-party-registration state is structurally unavailable, not zero.

Migration and economic context

Migration and economic modules use the source named in the record or methodology, including Census/ACS and other public statistical series where applicable. Time windows, inflation basis, and geographic crosswalks must remain attached to derived comparisons.

Officeholders and ideology

Current officeholder records combine official or maintained rosters with public biographical and legislative sources. DW-NOMINATE or similar ideology measures use their source's Congress/session and interpretation. An ideology score is a model coordinate, not a universal left-right truth.

Transformations

Common transformations include:

• identifier normalization;
• candidate and party normalization while retaining literal ballot labels;
• aggregation from precincts to place or district geography;
• reconciliation to certified county/state totals;
• boundary crosswalks;
• inflation or denominator normalization where stated;
• derived margin, swing, rank, similarity, and comparison metrics;
• static artifact generation and provenance manifests.

A transformation must never create the appearance of native precision that the source did not contain.

Licensing and attribution

Akashic's public license does not replace upstream terms. Before redistributing a combined dataset:

1. read the repository LICENSE and public ATTRIBUTION.txt;
2. retain source names and required notices;
3. check whether a source permits redistribution of raw files or only derived results;
4. preserve the retrieval date and transformation path;
5. avoid implying that an upstream agency endorses Akashic's interpretation.

Record-level verification

For a specific figure, prefer the provenance attached to the page/record or generated provenance manifest over this summary. Source coverage can change after a bake, and a single state-cycle may use a different file than the general source family.