Glossary.
Every domain term used on The Long Memory, defined. Geography acronyms, the archetype taxonomy, the demographic shorthand, and the methodology vocabulary — all in one place, all linkable.
On this page
- Geography — The legal and statistical units the federal government uses to describe American places.
- Politics & elections — The vocabulary of US presidential election analysis as we use it.
- Demographics — The statistical vocabulary we borrow from the Census Bureau and ASARB.
- Methodology — Terms specific to how The Long Memory is built.
Geography
The legal and statistical units the federal government uses to describe American places.
- FIPS code
Federal Information Processing Standards code identifying a US geographic area.
A five-digit FIPS code uniquely identifies a county (first two digits = state, last three = county). We use FIPS as the stable URL key for every county page (akashic.app/county/{FIPS}).
- CBSA
Core-Based Statistical Area — a metropolitan or micropolitan statistical area.
CBSAs are defined by the federal Office of Management and Budget as one or more counties sharing a labor market around an urban core of at least 10,000 people. The US has 918 CBSAs (387 Metropolitan, 531 Micropolitan as of 2023).
- MSA
Metropolitan Statistical Area — a CBSA with an urban core of 50,000+.
MSAs are the larger subset of CBSAs. "Pittsburgh, PA Metro Area" is an MSA; the smaller "Indiana, PA Micropolitan Area" is also a CBSA but not an MSA.
- DMA
Designated Market Area — a Nielsen-defined media market.
DMAs are Nielsen's 208 mutually exclusive television markets, used for advertising buys and political-ad targeting. Counties are assigned to a single DMA based on the dominant over-the-air signal.
- CD
Congressional District — one of the 435 US House districts (plus DC, PR, etc., as non-voting).
Congressional districts are redrawn each decade after the decennial census. The Long Memory currently serves the 118th Congress (2023–2025) boundaries; historical-boundary pages are on the roadmap.
- SLD-upper
State Legislative District, upper chamber (state senate).
Each state divides itself into legislative districts. The "upper" chamber is the state senate (or, in Nebraska, the unicameral legislature). District counts vary by state.
- SLD-lower
State Legislative District, lower chamber (state house or assembly).
The "lower" chamber in every bicameral state legislature. Typically smaller and more numerous than upper-chamber districts.
- Precinct
The smallest unit at which votes are reported — usually a single polling place.
A precinct (also "election district" in some states) is the level at which county election officials tabulate ballots. Precinct boundaries change frequently. The Long Memory uses 2024 precinct boundaries on its precinct maps, with vote totals disaggregated to the precinct level.
- Planning region
Connecticut's post-2022 replacement for counties.
In 2022 Connecticut formally replaced its eight legacy counties with nine Council of Government planning regions as its primary subdivision. The Long Memory uses planning-region geography for Connecticut and apportions pre-2022 county-level election totals to planning regions by 2020 town-level population.
- CDP
Census Designated Place — an unincorporated populated area the Census tracks as if it were a place.
CDPs let the Census report demographics for communities (typically suburban or unincorporated towns) that lack a municipal government. They appear in the long-tail /place/ surface alongside incorporated cities and towns.
Politics & elections
The vocabulary of US presidential election analysis as we use it.
- Margin
The two-party margin: (Democratic vote − Republican vote) ÷ total vote.
Expressed as a number in [−1, +1] or as a percentage. A margin of +0.10 means the Democratic candidate won by 10 percentage points. Third-party and write-in votes are included in the denominator.
- PVI
Partisan Voting Index — how a place votes relative to the nation.
PVI compares a place's presidential vote margin to the national popular-vote margin over recent cycles. "D+8" means the place is roughly eight points more Democratic than the country as a whole. The term originated with the Cook Political Report.
- Cook PVI
The original Cook Political Report PVI methodology.
Cook PVI uses a weighted average of the last two presidential elections (recent cycle weighted more heavily). Our methodology page documents how our equivalent metric is computed.
- Bellwether
A place that tracks the national winner closely, election after election.
In The Long Memory's archetype taxonomy, a bellwether is a county that has voted with the national winner in a high share of recent presidential elections, typically with narrow margins. See the methodology page for the exact threshold.
- Realigner
A county that voted one way for generations before swinging hard the other way.
Realigners are the archetype with the largest historical signal: a deep multi-decade lean to one party, followed by a multi-cycle swing in the opposite direction. McDowell County, WV (Democratic 1932–1996, Republican 2000–present) is the canonical example.
- Loyalist
A county that votes for the same party in nearly every election in living memory.
The Long Memory tags counties as Democratic loyalists or Republican loyalists based on their post-1932 voting record. The threshold and tie-breaks are documented on the methodology page.
- Urban anchor
A large urban center with a stable, dominant-party identity.
Urban anchors are the high-population counties that drive the partisan baseline of their region. Cook County (Chicago), Los Angeles County, and Bronx County are textbook Democratic anchors.
- Populist
A county with a large recent margin in a lower-income, working-class profile.
A modern archetype: places whose post-2008 voting pattern has tracked the populist insurgency on either left or right, paired with below-median household income and high non-degree share of adults.
- Recent convert
A county whose partisan identity has clearly shifted in the last sixteen years.
Distinct from a realigner in scale and depth: a recent convert may not yet have a multi-cycle baseline, but its 2008–2024 trajectory is unambiguous.
- Old Confederacy
A former Confederate State of America county; Democratic before 1960, Republican now.
The historical Solid South pattern: counties in the eleven CSA states that voted Democratic from Reconstruction through the mid-twentieth century, then swung to the Republican Party after 1964–1980.
- Western maverick
A Western-state county with an independent streak and high election-to-election volatility.
Western mavericks are counties (most commonly in MT, NV, NM, CO, AZ, ID) whose recent presidential vote swings more than 10 points between cycles, often crossing party lines.
- Frontier
A sparsely populated Western county with limited pre-1928 data.
Many Western counties were too sparsely populated to produce stable election results before the 1928 cycle. We tag them "frontier" rather than forcing them into a partisan archetype based on thin data.
- Tossup
A place whose most recent presidential margin is within two points.
A simple, recent-cycle archetype: the latest presidential election was decided by under two percentage points. Tossups are not necessarily competitive over the long run — many flip into Loyalist or Recent Convert in the next cycle.
- Archetype
A typology that classifies each county by its post-1932 voting trajectory.
The Long Memory assigns every US county to exactly one of twelve archetypes (Democratic loyalist, Republican loyalist, Realigner, Bellwether, Populist, Recent convert, Urban anchor, Old Confederacy, Western maverick, Frontier, Sparse, Tossup) based on a deterministic decision tree documented on the methodology page.
Demographics
The statistical vocabulary we borrow from the Census Bureau and ASARB.
- ACS
American Community Survey — the Census Bureau's rolling demographic survey.
ACS replaces the long-form decennial Census. The 5-year estimates (which we use) pool five years of survey responses to produce reliable estimates even for small geographies. Our data uses ACS 2024 5-year (reference period 2020–2024).
- ACS 5-year estimates
Five years of ACS responses pooled into one estimate.
The 5-year file is the only ACS product available for every county, no matter how small. Estimates carry sampling error; the Census Bureau publishes margins of error which we honor in our data validation pipeline.
- Median household income
The midpoint of household-income distribution for the place.
Reported in inflation-adjusted dollars for the ACS reference period. Half of households earn less; half earn more. Sensitive to ACS year — comparing across vintages requires adjusting for inflation.
- ASARB
Association of Statisticians of American Religious Bodies.
ASARB compiles the decennial US Religion Census, the most comprehensive county-level religious-adherence dataset. We use their 2020 release, bucketed into seven traditions for display.
- US Religion Census
Decennial county-level religious-adherence dataset compiled by ASARB.
The Religion Census reports the number of adherents per religious body per US county. We aggregate the ~250 reporting bodies into seven traditions: Baptist; Methodist; Pentecostal & Holiness; Catholic & Orthodox; Mainline Protestant; Other Christian; Non-Christian.
Methodology
Terms specific to how The Long Memory is built.
- The long memory
The 148-year presidential history (1876–2024) at the core of every page.
The product is named for the depth of the historical record: by going back to the post-Reconstruction era, we expose realignments, regional flips, and continuities that a recent-cycles-only view misses.
- Similar counties
The set of counties whose recent voting pattern most resembles a given county.
Computed as cosine similarity over the last-ten-election two-party margin vector. The result reflects political similarity over recent decades — not demographic or geographic similarity. Two counties with similar margin trajectories will appear similar even if they are on opposite coasts.
- Narrative
The deterministic prose paragraphs generated for each place page.
Every county page carries an editor-curated lead paragraph followed by paragraphs deterministically templated from the underlying election and demographic data. Same input, same output — no LLM in the runtime path.
- TIGER/Line
The Census Bureau's geographic-boundary shapefiles.
TIGER/Line files are the federal authoritative source for political and statistical boundary geometry. The Long Memory uses TIGER/Line 2024 county and precinct shapefiles, simplified with topojson-simplify for web delivery.
- ICPSR
Inter-university Consortium for Political and Social Research — historical election archive.
ICPSR (housed at the University of Michigan) maintains the canonical historical archive of US election results back to the early Republic. We use their county-level presidential series for pre-1916 cycles.
- MIT Election Lab
The contemporary county-level election results dataset, 1916–present.
The MIT Election Data and Science Lab publishes the most-cited modern county-level presidential election dataset. We use their series for 1916–2020 and the official state-certified returns for 2024.
- VEST
Voting and Election Science Team — precinct-level election results.
VEST is a project that assembles precinct-level vote totals from state and county sources. We use VEST's precinct data, aggregated to modern county boundaries, to produce the precinct maps on county pages.
- Cosine similarity
A measure of similarity between two vectors based on the angle between them.
For two vectors A and B, cosine similarity is (A · B) / (||A|| × ||B||), a value in [−1, +1]. Two vectors that point in the same direction score 1.0. We use it to find counties whose last-ten-election margin patterns most resemble each other.
See also
For the long-form discussion of how every term above is operationalized, see the methodology paper. For the project overview, see about. For what we’re building next, see the roadmap.