Political glossary: the terms Evidence uses to read an election.

Number of voters eligible to vote in a scope (country, state, abroad, municipality, zone or section) in an election, according to the Electoral Justice register. It is the base of turnout and abstention rates.

Registered voters by state

Number of eligible voters who voted. The turnout rate is turnout divided by registered voters of the same scope and round.

Formula comparecimento_pct = 100 * comparecimento / eleitorado

Turnout by state in 2026

Number of eligible voters who did not vote: registered voters minus turnout. In Brazil voting is compulsory for literate citizens aged 18 to 70 and optional for illiterate citizens, those over 70 and those aged 16 and 17 (Constitution, art. 14); those who miss it without a justification are subject to a fine.

Formula abstencao = eleitorado - comparecimento

Abstention in the 2026 case · Source

Scope of Brazilian voters who vote outside the country, in sections abroad. In TSE files it has the code ZZ and appears next to the 27 states; abroad, people only vote for President.

2026 election abroad

Each of the 26 states and the Federal District (Unidade da Federação). In TSE files each state has a two-letter lowercase code, such as sp, mg and ba.

The 27 states and abroad

Territorial unit of the Electoral Justice, under an electoral judge, grouping the sections of one or more municipalities.

Voting place of a group of voters, with one voting machine and one polling board. It is the smallest unit of the result: each section produces a polling station report.

Votes cast for candidates in a majoritarian election. Blank and null votes are not valid and do not count toward the absolute majority (Law 9,504/1997, art. 2).

Formula validos = comparecimento - brancos - nulos

Source

A vote in which the voter picks no candidate, using the BRANCO key on the machine. It is counted and published, but it is not a valid vote.

A vote for a number that matches no candidate, or annulled by the Electoral Justice. It is counted and published, but it is not a valid vote.

A candidate's votes divided by the valid votes of the same scope and round. It is the measure that decides a majoritarian election.

Formula pct = 100 * votos_candidato / validos

More than half of the valid votes. In the presidential election, whoever wins an absolute majority in round 1 is elected; otherwise there is a round 2 between the two most voted (Constitution, art. 77).

Source

Each vote of an election. Round 1 is on the first Sunday of October; round 2, when needed, on the last Sunday of October (Constitution, art. 77).

Results by round in 2026 · Source

Number that identifies a candidacy on the voting machine. For President it is the party number, so the same number may be used by different people in different cycles.

Comparison between cycles

Difference between the first and the second most voted in a round, in votes and in percentage points of valid votes.

Formula margem_pp = pct_primeiro - pct_segundo

Margin in the 2026 case

Unit of the difference between two percentages. From 45% to 48% is 3 percentage points, not 3%: in relative terms, the increase is 6.7%.

Formula diferenca_pp = pct_b - pct_a

Device on which the voter votes in the section. When voting ends, it prints the polling station report and records the result files, such as the digital vote record.

Document issued by each machine when voting ends, with the section's votes by candidate, blank and null votes and turnout. It is printed, posted at the section and sent for totalization, and it can be checked against the published result.

File recorded by the machine with each vote of the section in shuffled order, with no link to the voter. It allows the section to be recounted electronically.

Sum of the polling station reports of all sections by the Electoral Justice. Since 2020, totalization has been centralized at the TSE.

2026 counting timeline

Share of sections whose polling station report has entered the sum. A result with 90% of sections totalized is partial, not final.

Formula secoes_pct = 100 * secoes_totalizadas / secoes_total

Moment a scope (state or abroad) reaches 100% of sections totalized, recorded in the TSE coverage file. Evidence uses these moments in the counting timeline.

Closing time by state in 2026

TSE results app file (ending in -ab.json) with sections, registered voters, turnout and closing time of each state and abroad.

Source

TSE results app file (ending in -u.json) with the result of a scope in a round: votes by candidate, valid, blank and null votes, turnout and sections totalized.

Source

TSE portal with public election datasets, such as votes by municipality and zone. Evidence uses Open Data when a cycle is no longer in the results app, and marks the aggregate as derived.

2022 case · Source

Audit in which randomly drawn machines receive, on election day, votes from ballots filled in parallel, on camera, to check that the machine result matches the ballots.

Assurances recorded in the 2026 case

Moment in the count when the votes still to be totalized can no longer change who is ahead. It comes before complete totalization.

Act of the Electoral Justice that officially declares the election result after totalization.

Act of the Electoral Justice that hands the diploma to the person elected and qualifies them to take office. For President, the TSE does it.

Start of the term. Under Constitutional Amendment 111/2021, the term of the President elected in 2026 starts on January 5 of the year after the election.

Source

Change, in percentage points, of a candidate's share between two votes. In Evidence, swing between rounds compares the share between the two finalists in round 1 with the share of valid votes in round 2; swing between cycles compares the share of the same candidacy number in two elections.

Formula swing_pp = 100 * (parcela_t2 - parcela_t1_entre_finalistas)

Swing by state in 2026 · Source

A finalist's round 1 votes divided by the sum of both finalists' votes. It normalizes round 1 to compare it with round 2, where only the two compete.

Formula parcela_t1 = v1_A / (v1_A + v1_B)

A candidate's round 2 votes minus their round 1 votes, in the same scope.

Formula ganho = v2_A - v1_A

Evidence measure of how much the geographic make-up of the votes already totalized differs from that of the electorate at a moment of the count (total variation between expected and observed weight of each state). It helps avoid reading a partial count as the whole country.

Counting in the 2026 case

Share a candidate received from the votes totalized between two moments of the count only. It needs at least two collected moments; without them, it stays unknown.

Difference between the same measure in two elections, such as turnout in 2022 and 2026. Candidates, electorate and rules change between cycles: the difference describes two published results and measures no effect.

2022 x 2026 comparison

Hash function that turns a file into a string of 64 hexadecimal characters. Any change to the file changes the hash, so the recorded SHA-256 lets anyone check whether a file is the same one that was collected (NIST FIPS 180-4).

Sources of the 2026 case · Source

Evidence JSON file that lists every source collected for a case with URL, download date, size and SHA-256. Page generation refuses a missing file or one with a mismatched hash.

Manifest of the 2026 case

Reading of a result file at one moment, for one scope and round. A case may hold several snapshots of the same scope; pages use the latest.

Label every Evidence number carries to say where it comes from: observed, imported, derived, inferred, synthetic or unknown. The state is always visible next to the number.

Value read from an official source, with URL, date and SHA-256 recorded. Example: a candidate's votes in the TSE consolidated result file.

Value brought from a third party without auditing its method, such as a questionnaire result pasted by URL. It stands as a record of what the source said, not as a verified fact.

Value computed from observed or imported values, with the formula declared. Examples: turnout rate, margin and swing.

Value estimated by a model from incomplete evidence, with the uncertainty declared. It is not an observation and does not replace one.

Value generated by simulation or synthetic data, with no match to a real person or event. Example: Studio datasets.

Synthetic data studio

No collected source supports the value. Unknown does not mean false, and the reason is always recorded.

What is still unknown in the 2026 case

A checkable statement about a case, recorded in the ledger with the state the loaded evidence supports (confirmed, disputed, unsupported or unknown) and the rationale.

Claims ledger of the 2026 case

Record of a verification mechanism, such as an audit, a test or a software signature, with its source and what it does not support. An assurance covers what was verified, nothing more.

Assurances of the 2026 case

Axis of the Evidence political position instrument, with two poles and a scale from 0 to 100. There are 12 dimensions, read separately and never added into a label.

The 12 dimensions

Each end of a dimension. Pole A sits at score 0 and pole B at score 100. Pole names are conventions, not value judgments.

Instrument method