Sources and methods

The evidence layer behind every Akashic Edge result.

Akashic Edge aggregates, validates, and harmonizes U.S. election data from academic, governmental, and official election sources. This page is the working reference for coverage, source priority, processing rules, and citation standards across the platform.

Browse sources How to cite

Historical floor

1868

Earliest continuous presidential county coverage

Precinct geometry

163,926

2024 precincts with boundaries

Blocks

8M+

2020 block-level geographies in the warehouse

Page guide

Jump to the section you need

Methodology is often read under deadline. These anchors keep coverage, source provenance, processing rules, and citation formats easy to reach.

Coverage Sources Methodology Citations Infrastructure

Coverage overview

What the warehouse covers at a glance

These are the top-line dimensions of the current Akashic Edge election data footprint.

Election Results

19.4M+

rows across 51 state partitions

Elections Covered

224

1868–2024

Contests

151,488

President, Senate, Governor, House, State Leg

Counties

3,143

with geometry, demographics, and embeddings

Congressional Districts

435

118th Congress boundaries

Precincts (2024)

163,926

with full boundary geometry

Census Blocks

8M+

50 states, disaggregated election data

Geographic Entities

8.3M+

states, counties, CDs, SLDs, precincts, blocks

Office	Coverage	Geography	Notes
President	1868–2024	County + Precinct + Block	All 51 states. Block disaggregation for 50 states.
U.S. Senate	1908–2024	County	All regular and special elections.
Governor	1865–2025	County	Includes off-year elections such as Virginia and New Jersey.
U.S. House	1976–2024	Congressional district	MEDSL source. 10,869 contests.
CD Presidential	2008–2024	Congressional district (118th)	435 CDs. 2008–2020 projected onto current boundaries.
State Senate	1968–2023	State legislative district	Klarner dataset, all 50 states.
State House	1968–2023	State legislative district	Klarner dataset, all 50 states (excluding Nebraska unicameral).

Source inventory

Primary source families

Akashic favors official and academically maintained datasets, then applies a clear priority order when multiple sources overlap.

Historical data

Pre-2000 federal and state-level historical returns.

Algara-Amlani Historical Election Dataset

Carlos Algara & Daniel Amlani (UC Davis / University of Houston)

President, Senate, Governor: 1868–2020County-level (3,100+ counties)Academic (CC-BY)

1.015M rows. Primary source for historical county data.

ICPSR 0001 — U.S. Historical Election Returns

Inter-university Consortium for Political and Social Research

Source

1824–1968 (codebooks only — data access restricted)County-levelICPSR member institutions

Used for validation cross-referencing.

ICPSR 0013 — General Election Data

ICPSR

Source

1950–1990County-level, all federal + GovernorICPSR member institutions

Modern elections

County, district, and precinct-era election results.

MIT Election Data + Science Lab (MEDSL)

Massachusetts Institute of Technology

Source

County presidential 2000–2024; U.S. House/Senate (county-level)County + partial precinctCC-BY

12.5K county pres rows + 33.7K House rows.

VEST — Voting and Election Science Team

Harvard Dataverse / VEST

Source

Precinct-level results 2016–2020 (46/51 states)Precinct boundaries + results + block crosswalksCC-BY

429K precinct results with geometry.

New York Times 2024 Precinct Results

The New York Times

2024 presidential (50/51 states, AK unavailable)163,926 precincts with full boundary geometryDerived from official certified results

481,541 results (D/R/OTH) with geometry.

Klarner State Legislative Dataset

Carl Klarner (Indiana State University)

State Senate + State House: 1968–2023State legislative districtsAcademic

347K results across all 50 states.

State Secretaries of State

Official state election authorities

Certified results (varies by state)Precinct or county-levelPublic domain

Used for Alaska (DoE certified), 2024 manual backfills.

Census and geometry

Population weights, demographics, and boundary infrastructure.

Decennial Census (PL 94-171)

U.S. Census Bureau

Source

Total population + race at block level (2020)8M+ census blocks across 50 statesPublic domain

Used as population weights for block-level disaggregation.

American Community Survey (ACS) 5-Year

U.S. Census Bureau

Source

Socioeconomic data (2024 5-year estimates)County + congressional districtPublic domain

3,639 ancestry rows (European, MENA, African, Latino, Asian, AIAN, NHPI).

TIGER/Line Shapefiles

U.S. Census Bureau

Source

All census geographies (2020 vintage)States, counties, CDs, blocks, precinctsPublic domain

PostGIS boundaries for 8.3M geographic entities.

Processing rules

Methodology that governs outputs

These conventions define how Akashic computes margins, projects results, resolves overlaps, and validates the warehouse.

Margin convention

All partisan margins are calculated as the Democratic vote total minus the Republican vote total, divided by total votes cast, multiplied by 100. This is share of all votes cast, not two-party normalized margin.

Block-level disaggregation

For precinct-to-block disaggregation, Akashic uses a population-weighted centroid-in-polygon method. Each block receives its share of precinct votes based on 2020 PL 94-171 population, and Hare quota rounding reconciles totals back to certified precinct results.

CD presidential projection

Congressional district presidential results for 2008–2020 are projected onto 118th Congress boundaries using area-weighted spatial crosswalks between historical precincts and current district geometry. 2024 uses actual 118th boundaries.

Winner determination

The `is_winner` flag is populated algorithmically by selecting the highest vote total at the contest level. This matches certified outcomes in more than 99.9% of cases.

Fusion voting

In states with fusion voting, candidate vote totals are consolidated across ballot lines so cross-endorsed candidates appear with a unified total rather than duplicate party-line rows.

Data validation

Automated validation checks vote-total reconciliation, geographic completeness, margin accuracy, and cross-source consistency on a daily cadence. Anomalies are logged for manual review.

County similarity

County similarity uses 32-dimensional embeddings combining demographic, political, and composite features. HNSW indexing supports fast nearest-neighbor lookup, and K-Means clustering groups counties into twelve archetypes.

Canonical formula

Margin convention

Akashic uses share of all votes cast, not two-party normalized margin.

margin = (Democratic votes - Republican votes) / total votes * 100

Priority order

Source priority matrix

When multiple source families cover the same period and geography, Akashic resolves conflicts using this order.

Period	County source	Precinct source
1868–1949	Algara-Amlani → ICPSR 0001	N/A
1950–1968	Algara-Amlani → ICPSR 0013	N/A
1969–1999	Algara-Amlani → State SOS	N/A
2000–2015	MEDSL → Algara-Amlani → State SOS	State SOS (partial)
2016–present	MEDSL (county)	VEST (2020) → NYT (2024) → State SOS

Citation rules

How to cite Akashic Edge data

Include the office, geographic level, and date range when citing a specific subset. That keeps downstream citations precise and reproducible.

APA

Akashic Edge. (2026). U.S. Election Data [Dataset]. Retrieved from https://akashicedge.com

Chicago / Turabian

Akashic Edge. "U.S. Election Data." Accessed [date]. https://akashicedge.com.

BibTeX

@misc{akashicedge2026,
  title  = {U.S. Election Data},
  author = {Akashic Edge},
  year   = {2026},
  url    = {https://akashicedge.com},
  note   = {Accessed: [date]}
}

Citation note

For subset citations, include the office, geography, and covered date range. Example: Presidential county-level results, 1868–2024, via Akashic Edge.

Methodology questions

Researchers, librarians, and data teams can reach out for institutional access, bulk data requests, or source clarification.

team@akashicedge.com

Technical substrate

Infrastructure that supports the data layer

The methodology page should also make the storage and query substrate legible for institutional evaluators.

Database

PostgreSQL 16 + PostGIS 3.4 + pgvector + pg_trgm

Hosting

Vercel (edge) + Neon (serverless Postgres)

Partitioning

51 state partitions for election_results

Spatial index

GiST on geometry columns

Similarity index

HNSW (pgvector) for county embeddings

Vector tiles

21,146 .pbf files (states z0-8, counties z4-10, CDs z3-10)

Materialized views

4 MVs (19.4M + 18.3M + 6.6M + 510K rows)

Update cadence

Continuous during election seasons; daily validation