Methodology
How we compile and maintain comprehensive data on state elections — and only state elections — covering nearly 8,000 elected offices across all 50 states.
How we compile and maintain comprehensive data on state elections — and only state elections — covering nearly 8,000 elected offices across all 50 states.
This site covers state elections — not federal, not local. Governors, statewide executive offices, state legislatures, and statewide ballot measures, in all 50 states. You will not find congressional races, presidential contests, or city councils here, and that is deliberate: state government is what we know best, and staying inside that lane is what lets the data go this deep. The database centers on the current election cycle and includes historical elections back to 1948, with depth varying by state and era.
One acknowledgment before the details: a database of this size contains mistakes. Some are ours; some arrive in the official record itself, which can be wrong, incomplete, or corrupted in transmission. We audit continuously and correct at the source, and if you spot something wrong, tell us — a correction lands in the database itself, so the fix reaches every page that displayed the number.
We track every state-level office that voters elect directly. The counts below reflect the states where each office is filled by election; elsewhere it is appointed, runs on a joint ticket, or does not exist.
Official state government records are the backbone of this database. Secretaries of State, state election boards, and legislative clerks are the source of record for candidate filings, vote totals, certifications, and officeholder rosters, and we go to them directly rather than to a compilation wherever we can. Where sources disagree, the official record controls.
We also lean on — and genuinely admire — the work of several organizations in this space. Some produced analytical datasets that appear on this site with credit attached to every figure; others serve as the reference we check our own records against. All of them are worth your time in their own right, and the table notes what each contributes here.
Forecast ratings are a different thing entirely: we do not forecast, and the ratings displayed on this site are the editorial work of the forecasters named in the Forecast Ratings section below.
| Source | Data Used | Role |
|---|---|---|
| Secretaries of State and state election boards | Candidate filings, candidate status, election results, official canvasses and certifications | Primary source (always preferred) |
| State legislature websites and chamber clerks | Current member rosters, vacancy notices, district and term information | Official source |
| StateNavigate | 2024 presidential results disaggregated to state legislative districts. Now credited on nine Maine House districts; previously the source of our 2024 baselines site-wide, and still an independent cross-check on our own 2024 figures | District partisan baselines (2024) |
| Federal Election Commission | Certified presidential results by state — the statewide 2024 margins shown on state and office pages, and the gate every one of our district calculations must reproduce | Official record |
| Daily Kos Elections / The Downballot | 2012 and 2016 presidential results by state legislative district (47 states each), plus 2020 for 13 states, retained as an independent check on our own 2020 build | District partisan baselines (2012, 2016, 2020) |
| VEST (Voting and Election Science Team) — Harvard Dataverse | Precinct-level results with boundaries: 2020 (50 states) and 2016 (AL, AR, MS) | District partisan baselines (2016, 2020) |
| Harvard Election Data Archive / Stanford precinct project | 2008 precinct-level returns geo-referenced to Census voting districts | District partisan baselines (2008) |
| MIT Election Data + Science Lab (MEDSL) | Precinct-level returns used in a handful of state builds where the official file needed a second copy | District partisan baselines (supporting) |
| U.S. Census Bureau | TIGER/Line legislative district and voting district boundaries (2010 through current vintages); 2020 census block populations (PL 94-171) for split-precinct allocation | District partisan baselines (all years) |
| National Conference of State Legislatures | Chamber control history, legislative structure and session information | Check-and-verify reference for our own records |
| OpenElections | Precinct-level historical election results, including compilations of official county returns used in several of our 2008 and 2012 baseline builds | Historical reference; district partisan baselines (2008, 2012) |
| US Election Atlas | Historical statewide results by office, back to the 19th century | Historical reference |
| OpenStates | Current state legislator rosters | Check-and-verify reference for our rosters |
| Wikipedia | Historical officeholder lists and biographical detail | Check-and-verify reference for historical records |
| Ballotpedia | Historical election and candidate records imported during this site's early development, which remain credited in the database | Historical records |
Where a number on this site is another organization's analysis rather than a government record, that organization is named in the table above and, where the number is displayed prominently, alongside the number itself.
Election results move through three stages on this site, and every race carries its current stage.
A race moves to Called one of three ways. Most commonly it clears an arithmetic threshold: the leader's margin is at least 2 percentage points and 100 votes, at least 95% of precincts have reported, the margin exceeds twice the estimated outstanding ballots, and the race is not inside the state's statutory recount window. Races that do not clear that bar can still be called on a documented real-world signal, such as a concession, a declined recount, or a call by a major decision desk. Races with a single candidate on the ballot are called automatically once the election has been held.
Moving a race to Certified requires an official certification record from the state, retrieved at the time we make the change. Consistency with earlier unofficial numbers is not sufficient. Because certification deadlines vary widely by state, a race can sit at Called for weeks after election night. For more on that process, see our guide to certification and our guide to recounts.
Status changes are logged, and the rare downgrade — for example, a call we reverse because the underlying source turned out not to support it — is recorded the same way.
Winning an election and holding an office are different facts, and we track both. Every seat carries a tenure history — who held it, from when to when, and how each term began and ended — built from more than 60,000 individual term records. That history is what powers the officeholder timeline on district and office pages, and it is why a seat can show a vacancy, an appointee, or a mid-term successor rather than simply the last election's winner.
The distinctions we record are the ones that change how a seat should be read:
The timing here follows the real institutions rather than the news cycle: a general election winner does not appear as the officeholder until the results are certified and the term begins, because the incumbent holds the seat until then. Special election winners are seated on the faster timelines those elections actually use.
An officeholder has two party facts, and they are not always the same one:
For maps, charts, and partisan counts, we use caucus affiliation where it differs from ballot party. The distinction is load-bearing in several states:
Because of this, our chamber composition figures can differ by a seat or two from counts published elsewhere that use raw party registration. That difference is deliberate.
A government trifecta exists when one party controls the governorship and both chambers of the state legislature. We classify each state as:
Chamber control is determined by the majority caucus, accounting for vacancies, tied chambers, and power-sharing arrangements case by case. See our guide to trifectas.
Forecast ratings on this site cover governor races and come from Sabato's Crystal Ball (University of Virginia Center for Politics), the Cook Political Report, and Pluribus News. All three use the same conventional seven-point scale.
| Rating | Probability for Leading Party | Description |
|---|---|---|
| Solid D / Solid R | >95% | Not seriously contested; the outcome is a near-certainty |
| Likely D / Likely R | ~80–95% | Clear favorite with a credible path for the underdog |
| Lean D / Lean R | ~65–80% | Genuine contest; one party has a meaningful but not commanding advantage |
| Toss-up | ~45–55% | No clear advantage for either party; essentially a coin flip |
Where more than one forecaster has rated a race, the rating shown on our maps is a consensus derived from all available ratings rather than any single forecaster's call. Individual ratings are shown on the race detail page with each forecaster named.
The probability ranges above reflect conventional interpretations of each tier. The forecasters publish qualitative ratings, not explicit win probabilities, and their ratings should be read as expert judgment about the competitive landscape rather than statistical predictions.
We do not produce forecast ratings of our own, and we do not carry ratings for individual legislative districts. Forecast ratings are the one place on this site where a displayed figure is an outside organization's editorial judgment rather than a government record, which is why each rating is labeled with the name of the forecaster who issued it.
A district number is a name that different maps reuse. "House District 12" in 2008 and "House District 12" today can be entirely different territory, and treating them as one thing quietly corrupts everything downstream — election history, officeholder tenure, and every baseline. So districts in our database belong to a redistricting era, and each election stays attached to the district as it existed when that election was run.
The eras are not just the decennial cycles. Court-ordered redraws in mid-decade — a regular feature of the last two decades of redistricting litigation — get their own era where the lines genuinely changed, so a race run on remedial lines is not measured against the map it replaced.
District pages make this visible rather than hiding it: the district map offers the current lines and the earlier eras we hold boundaries for, drawn from Census TIGER/Line files of each vintage, and the page's redistricting timeline marks where the lines changed and what the change did to the district's partisan lean. Where a state has renamed or renumbered its districts across eras, we preserve each era's own naming rather than forcing old results into current labels.
A district's presidential result is the standard partisan baseline used by election analysts. It describes the underlying lean of the territory itself, independent of who happens to be on the ballot, which is what makes it useful for judging whether a legislative race is genuinely competitive or whether a candidate over-performed the ground they were running on.
We hold five presidential elections mapped to state legislative districts, giving the large majority of the races on this site a baseline computed for the territory they were fought on. Every figure carries the name of the organization that produced it, and that credit is attached to the number in our database rather than kept in a list on this page, so it travels with the data wherever it is displayed.
| Presidential year | District lines | Coverage | Produced by |
|---|---|---|---|
| 2024 | Current (2022 cycle) | All 50 states | Calculated by MultiState from official state and county records. Exception: nine Maine House districts come from StateNavigate. Statewide margins are the FEC-certified official results |
| 2020 | Both: pre-2022 maps and current lines | All 50 states, on each set of lines | Calculated by MultiState from VEST precinct data. Daily Kos Elections' 13-state 2020 dataset is retained alongside as an independent check |
| 2016 | Pre-2022 maps | All 50 states | Daily Kos Elections for 47 states; calculated by MultiState (from VEST) for Alabama, Arkansas, and Mississippi |
| 2012 | Pre-2022 maps | All 50 states | Daily Kos Elections for 47 states; calculated by MultiState for Alabama, Maryland, and Mississippi |
| 2008 | 2000s maps | 40 states | Calculated by MultiState from archival precinct records (see below) |
We calculated the 2024 figures — the "2024 Pres" number on every current-cycle race card, the partisan lean on district pages, the district comparison feature, and our count of crossover districts — state by state from official precinct-level sources: state election agencies' precinct results and boundary files, the states' own legislative returns, and, where a state does not publish usable precinct data, certified county records. A state's figures are accepted only when they reproduce the official certified statewide result, and the statewide margins themselves are those FEC-certified figures, taken directly from the official record. One piece still comes from StateNavigate: nine Maine House districts we could not calculate, because their split towns' presidential vote was never tabulated by district portion. StateNavigate's dataset was also the original source of our 2024 baselines site-wide, and it remains a valuable independent check: in states where we can compare, the two independently-built datasets typically agree to a tenth of a point.
The 2012 and 2016 figures for most states come from Daily Kos Elections, whose Statewide Results by LD dataset (Jeff Singer, Daniel Donner, Stephen Wolf, and David Nir) is the standard reference for the pre-2022 map era. We calculated the three states each year that their dataset does not cover.
We calculated the 2020 figures from the VEST (Voting and Election Science Team) precinct dataset for all 50 states — twice, once on the maps 2020 was actually run under and once on today's lines, so every current district has a second baseline alongside 2024. Daily Kos Elections' 13-state 2020 dataset was deliberately left in place as an independent check on our method: across 1,060 districts in common, the median difference is 0.03 points.
Presidential results are reported by precinct, not by legislative district, so a district-level figure is always an allocation of precinct returns onto district boundaries. That carries a consequence worth stating plainly: a baseline is only valid for the lines it was computed on. When a state redistricts, the territory changes, and the old number stops describing the new district.
So we pair every race with the baseline calculated on the map that race was actually run under, and we choose the map era before choosing the year. A 2018 North Carolina race is measured against 2016 results computed on the 2018 lines, not the 2012 lines that district previously used and not today's. Where a map was adopted after the last presidential election, we use the earliest presidential election held on those same lines, which is a later election than the race but the correct ground.
The corollary matters too: the difference between a district's 2016 figure and its 2024 figure is not a measure of how that area shifted, because the two describe different geography. We do not present it as one.
Where a state redrew its maps mid-decade, the same presidential election can be calculated onto both the old and the new boundaries. Holding the election constant isolates the effect of the redraw by itself: North Carolina's 1st Senate district is R+22.5 on the 2012 map and R+12.3 on the 2018 map, using identical 2016 votes. Daily Kos Elections did this for the mid-decade redraws in Alaska, North Carolina, and Virginia, and our own 2020 build does it for every state across the 2022 redistricting. Where we have both, the district page's redistricting timeline shows that change directly.
We calculated most of this dataset ourselves, because no published dataset covered it: all of 2024, all of 2020 on both map eras, 2008 for 40 states, and the states each year that the Daily Kos Elections series does not include. Around 25,000 of the roughly 40,000 district-year figures on this site were calculated in-house from primary sources, and the standing goal is to independently calculate or verify every year we hold, extending coverage as far as the surviving records allow.
The 2024 layer was calculated one state at a time from what each state actually publishes: precinct results joined to the state's own boundary files, statewide files where precincts carry their legislative district as a column, and — the best case, available in more than a dozen states — the state's own legislative-contest returns, which constitute an authoritative precinct-to-district assignment counted from real ballots rather than estimated from geometry. Where a state publishes none of these, we used certified county records.
The 2008 layer, on the 2000s maps, exists nowhere else in any form — Daily Kos Elections' legislative series begins in 2012, and their 2000–2008 work covers congressional districts only. Without it, every race of the 2000s map era, including the 2010 wave cycle, would have no baseline at all. We calculated it from precinct returns in the Harvard Election Data Archive, official state election-history databases, OpenElections compilations of certified county returns, and state archives, matched to Census voting-district geography and the 2010-vintage TIGER/Line district boundaries — the maps those elections were actually run under.
Every build passes the same gates before it is written. A state is accepted only when its aggregated precinct data reproduces the officially certified statewide presidential result — most of our states match to the hundredth of a point, and many to the exact vote — and only when its districts join our records completely. Where the surviving records genuinely cannot support assigning a share of a state's vote to districts, we exclude those districts rather than publish a smeared estimate; a missing baseline is honest, a flattened one is not. That is why 2008 covers 40 states rather than 50, and why a few districts within covered states show no figure.
Most precincts sit entirely inside one legislative district, and their votes are assigned to it exactly. Only precincts that genuinely straddle a boundary need splitting, and those we divide by the 2020 census block population that falls on each side — the same method the major published datasets use — with land area as a fallback only where a split piece contains no people at all. Where a state's own legislative returns provide a counted assignment, nothing is split at all.
Split-precinct allocation is still the weakest link in any district-level presidential figure, ours included, so we keep its footprint measured and small: in a typical state, well over 90% of the vote falls in precincts that nest cleanly inside a single district, and the estimated share is confined to the remainder.
Independent agreement is the strongest evidence a method can offer, and we check it wherever an independent dataset exists. Our 2020 build reproduces Daily Kos Elections' figures for their 13 states to a median difference of 0.03 points across 1,060 districts, with no partisan lean in the differences. Our 2024 build independently arrives at StateNavigate's figures in most states to within rounding. We also cross-check district turnout against each district's own legislative-contest turnout — presidential votes in a district cannot meaningfully fall below the district's own down-ballot count — and, for historical years, verify that a district's movement between adjacent elections on the same lines tracks its state's overall swing.
Presidential percentages are a share of the total presidential vote cast in the district, with third-party candidates in the denominator. Margins are stated as D+X or R+X. Because every figure is an allocation of precinct returns onto district boundaries rather than an officially reported district total, small differences between sources are expected and normal.
We show a presidential baseline on general elections only. A party primary is decided within one party's electorate, where a districtwide presidential margin says little of use.
Coverage varies by year and by state, and a race we have no matching baseline for shows none. We would rather show nothing than borrow a number computed for different territory.
Election data follows a seasonal rhythm, and our update schedule matches the electoral calendar.
On election night and during active counting, current-cycle results on district pages are read live from our database rather than from a cached file, so the page reflects the most recent import without waiting on a site rebuild.
A database this size fails quietly if it is not audited loudly. A suite of automated integrity checks runs against the full database — not just new imports — looking for the failure modes we have actually encountered: officeholder terms that overlap on one seat, winners without terms and terms without wins, duplicate candidate records, elections attached to the wrong era's district, and result statuses that moved in a direction they should not.
One person is one record. Official filings use legal names while ballots and news coverage use common ones, and the same person may file as "Charles Edward Hufstetler III" in one record and appear as "Chuck Hufstetler" in another. Every candidate import runs through a matching step built for exactly this — nicknames, middle names, suffixes, and initials — so a candidate's full history across cycles, districts, and offices attaches to a single record. Where a match is genuinely ambiguous (a son filing for his father's seat produces identical-looking signals), we verify against public records rather than merge on similarity.
Vote totals get checked against more than their own source. On election night, imported totals are compared across independent feeds and against the state's own reporting before and after loading; a systematic discrepancy blocks the import rather than shipping. Certified totals replace unofficial ones only from a certification record retrieved at the time of the change.
Not every district elects one member, and the states that elect several do it in structurally different ways. Getting this right matters, because a naive reading of the returns attributes wins to the wrong people.
We track each seat individually, and where a state has changed its structure over time, historical results stay attached to the district as it existed then. See our guide to districts.
Nebraska's unicameral legislature holds officially nonpartisan elections, and candidates appear on the ballot without party labels. Members' affiliations are well known and publicly reported, and we track them as caucus affiliation for analytical purposes while recording their formal party as nonpartisan.
California and Washington use a top-two primary in which all candidates appear on one ballot regardless of party and the two highest finishers advance, which can produce a general election between two candidates of the same party. Alaska uses a top-four version. Louisiana holds an all-candidate election in which a majority winner is elected outright. We record these as a single primary contest rather than separate party primaries.
Maine uses ranked-choice voting for state primaries, and Alaska uses it for the general election following its top-four primary. In both, the first-round leader can differ from the eventual winner, so we do not declare a result from first-choice votes alone. Our data reflects the final tabulation after all rounds, and we record the round-by-round detail where the state publishes it.
We would rather hear about an error than have it sit. If you find something wrong, or you believe we have used your work without proper credit, contact us and we will look into it and correct the record. Corrections are made in the database itself, so the fix propagates to every page that displayed the number.
MultiState is the leading state and local government relations firm. We help organizations navigate the complex landscape of state policy through legislative tracking, regulatory compliance, and strategic consulting across all 50 states.
MultiState Elections is a public resource that reflects our deep expertise in state government. For organizations that need to go beyond tracking elections to tracking legislation, regulations, and policy, explore our full range of services.