Municipality by municipality: what changed from the 2022 first round to the 2026 one

Note to the reader: in this text, "the PL candidate" is Jair Bolsonaro in 2018 and 2022 and Flávio Bolsonaro in 2026; "the PT candidate" is Fernando Haddad in 2018 and Lula in 2022 and 2026. Comparing candidacies is not comparing people. "pp" means percentage point. All vote percentages are of valid votes (excluding blank and spoiled ballots) unless stated otherwise.

What happened

In the first round on Oct. 4, 2026, Flávio Bolsonaro (PL) received 56,104,503 votes, 47.03% of the valid votes. Lula (PT) received 53,879,538, or 45.16%. The gap was 2,224,965 votes, or 1.87 percentage points. Neither passed 50% of the valid votes, and the two face each other in the runoff on Oct. 25. The other ten candidates together took 7.81% (9,316,747 votes); the three largest were Augusto Cury (Avante), with 2.89%, Renan Santos (Missão), with 2.24%, and Ronaldo Caiado (PSD), with 2.18%. (Source: TSE, the Superior Electoral Court, votes by candidate and municipality/zone, file generated on Oct. 21, 2026.)

For scale, the same stage in the two previous elections. In 2018, Jair Bolsonaro (PSL) received 46.03% of the valid votes (49,277,010 votes) and Fernando Haddad (PT), 29.28% (31,342,051): a gap of 16.8 pp. In 2022, Lula received 48.43% (57,259,504) and Jair Bolsonaro (PL), 43.20% (51,072,345): a 5.2 pp gap in Lula's favor. In 2026, the gap is 1.9 pp in favor of the PL candidate.

The vote count from 2022 to 2026 shows where the change came from. The PL candidate has 5.03 million more votes; Lula, 3.38 million fewer. The other candidates together have 0.58 million fewer, blank votes are up 0.34 million, spoiled votes are up 0.19 million, and abstention is up 0.70 million. The electorate grew by 2.29 million (from 156.45 million to 158.75 million eligible voters).

One fact that often goes unnoticed: almost all the change in Lula's vote comes from his share of the valid votes, and almost none from turnout. Decomposing Lula's 3.38 million fewer votes, electorate growth contributed +0.81 million, turnout −0.10 million, the proportion of valid votes among those who showed up −0.21 million, and Lula's share of the valid votes −3.88 million. For the PL candidate, the same calculation gives +0.78, −0.09, −0.20 and +4.55 million. Abstention, then, does not explain the difference between the two years (more in finding 3).

On the map of the states. According to TSE data, Lula received the most valid votes in 12 federative units: the nine states of the Northeast, plus Pará, Amazonas and Amapá. The PL candidate led in 15: 14 states and the Federal District. In 2022, Lula led in 14 units and Jair Bolsonaro in 13. Minas Gerais and Tocantins went from Lula to the PL candidate. Congresso em Foco, a news outlet, arrived at the same count of 12 and 15 (see section 5).

The tightest result in the country was in Amapá: Lula received 212,503 votes and the PL candidate, 212,278. That is a difference of 225 votes, out of 464,858 valid votes. In Minas Gerais, the margin was 4.9 pp; in Amazonas, 3.2 pp; in Pará, 5.4 pp; in Tocantins, 7.0 pp.

On the map of the municipalities. Lula received the most votes in 2,665 of the 5,570 municipalities, the PL candidate in 2,905. In 2022, the first-round split was 3,379 for Lula and 2,191 for Jair Bolsonaro. Among the 41 municipalities with more than 500,000 inhabitants (2022 Census), Lula led in 15 and the PL candidate in 26.

Blank votes, spoiled votes and abstention

The numbers. In the 2026 first round, 33,469,244 voters did not show up: 21.08% of the 158,745,502 eligible voters (voting is compulsory in Brazil, but the penalty for not voting is small). It is the highest abstention in a presidential first round since 1998 (21.49%). In the midterm elections of the series we compiled, it was 17.74% in 2002, 16.75% in 2006, 18.12% in 2010, 19.39% in 2014, 20.33% in 2018 and 20.95% in 2022. The difference from 2022 is 0.13 pp, or about 700,000 more voters.

Among those who turned out (125,275,835 people), 1.84% cast a blank vote and 2.93% spoiled their ballot, 4.77% in total. In 2022, the figures were 1.59% and 2.82%, 4.41% in total; in 2018, 2.65% and 6.14%, 8.79%. The drop from 2018 to 2022 stands out: spoiled votes went from 7.2 million to 3.5 million. We did not find a source that explains this with data, so we do not propose an explanation. Adding abstention, blank and spoiled votes, 24.85% of eligible voters did not choose any candidate in 2026, against 24.43% in 2022 and 27.33% in 2018.

Where they are highest. Abstention in 2026 was highest in Mato Grosso do Sul (23.87%), Rio de Janeiro (23.32%), Minas Gerais (22.81%), São Paulo (22.54%) and Mato Grosso (22.20%). It was lowest in Ceará (15.52%), Roraima (17.12%), Piauí (17.15%), Sergipe (17.30%) and Paraíba (17.38%). Abroad, where 916,534 voters were eligible, abstention was 62.66%.

Among the state capitals, Rio de Janeiro had the highest abstention (26.12%) and Fortaleza the lowest (13.02%). Among municipalities, the highest rates are in Minas Gerais: Rio Vermelho, at 40.09%, tops the list, and the first 5 places in the ranking are all in Minas Gerais. By size, abstention is similar across all brackets, from 20.3% (municipalities of up to 5,000 inhabitants) to 21.3% (20,000 to 50,000).

Compared with 2022, abstention fell in 13 federative units and rose in 14 (not counting voters abroad, where it rose from 56.19% to 62.66%). It fell most in Rondônia (−2.94 pp), Maranhão (−2.47 pp) and Pará (−2.12 pp). It rose most in Mato Grosso do Sul (+1.78 pp), Rio Grande do Sul (+1.57 pp) and the Federal District (+1.36 pp).

Blank and spoiled votes together (as a % of those who turned out) were most frequent in Paraíba (6.34%), Alagoas (6.17%), São Paulo (5.86%), Bahia (5.45%), Pernambuco (5.25%) and Minas Gerais (5.23%), and least frequent in Roraima (2.17%), Rondônia (2.32%) and Amapá (2.46%). The biggest increases compared with 2022 were in Pará (from 2.28% to 4.01%), Alagoas (from 4.53% to 6.17%) and Ceará (from 3.54% to 4.95%).

What other sources say. Each reading below belongs to whoever makes it, not to us.

• Fatigue with polarization. Political scientist Hely Ferreira, interviewed by Diário de Pernambuco, attributes part of the abstention to weariness with polarization and the lack of a competitive "third way," which would have led dissatisfied voters to stay home (Diário de Pernambuco). In 2026 there were three candidates between 2.2% and 2.9% of the valid votes; we did not test the hypothesis.
• Profile of those who stay home. According to Arthur Wittenberg, of Ibmec-DF, the right tends to miss fewer votes but is harder to mobilize, while the left, on average lower in income and education, abstains more but responds more to mobilization. Leandro Gabiati, of the consultancy Dominium, observes that polarization can mobilize people who think a few votes will decide the race. The same article covers factors that the outlet itself presents, without attributing them to experts, such as the aging of the electorate, change of residence without transferring the voter registration, and the low fine (Gazeta do Povo). Before the election, the article projected abstention above 22% in the first round (the final figure is 21.08%) and, according to Gabiati, 25% when adding blank and spoiled votes (we calculate 24.85% of eligible voters).
• Disinterest and alienation. The study "Electoral Alienation in Democratic Brazil" ("Alienação Eleitoral no Brasil Democrático"), by Instituto Votorantim, released in 2022 by Estadão/Broadcast, adds up abstention and blank and spoiled votes and links the increase to corruption scandals and the perception that the vote has no value (Rafael Gioielli, of the institute), and to frustration with democracy and party fragmentation (Nara Pavão, of UFPE). The study also notes that voters with a college education vote more and that young people up to 24 abstain more (O Povo/Agência Estado). The numbers the study reports for "alienation" (18% in 2006 and 25% in 2018) do not match the sum we calculate for the same concept over eligible voters (23.76% and 27.33%). We do not know their calculation base (guess: a different definition of the denominator), so we did not use the study's numbers.
• Two academic interpretations. Marques and Souza (UFMG), in a paper presented in 2020 at the ANPOCS meeting, offer two possible readings of "electoral alienation": protest by voters dissatisfied with politicians, or a shift in values and schooling that would lead mainly young people to prefer other forms of participation. The paper does not test these hypotheses directly (TSE Digital Library).
• Institutional factors. Timothy Power and J. Timmons Roberts, in a 1997 article in the journal Estudos Eleitorais, studied legislative elections from 1945 to 1990 and argue that, besides protest and socioeconomic level, institutional design (a complex ballot, many candidates and parties) helps explain invalid votes (TSE Digital Library). The study is from another era and about legislative elections; it serves as a conceptual reference, not as an explanation of 2026.
• Opinion column. Columnist Cleyton Monte, in a column published on Oct. 15, 2026 in O Povo, states that abstention is higher among people with lower income and education and that this represents a challenge for Lula. The column presents no numbers or sources for this claim (O Povo).

What our data say. With TSE data, it is not possible to know the income or education of those who abstained. What can be measured is the municipality. Comparing municipalities within the same state, where abstention rose more than the state average, Lula's share fell less (correlation of 0.14), and the PL candidate's rose less (−0.13). The relationship is weak and has the opposite sign from the national correlation (−0.24), a case of reversal caused by mixing states, which we explain in finding 3.

State by state and city by city

The overview. From 2022 to 2026, Lula's share of the valid votes fell in 25 of the 27 states. The exceptions were the Federal District (+1.26 pp) and Amapá (+0.04 pp, practically stable). The PL candidate's share rose in 26 states; the exception was the Federal District (−0.34 pp). Abroad, Lula's share rose 0.39 pp and the PL candidate's, 1.88 pp.

The biggest drops in Lula's share were in Goiás (−8.46 pp), Tocantins (−6.97 pp), Rio Grande do Sul (−6.56 pp), Mato Grosso (−5.21 pp), Minas Gerais (−4.96 pp), Maranhão (−4.85 pp), Paraná (−4.79 pp), Santa Catarina (−4.49 pp) and Mato Grosso do Sul (−4.36 pp). The biggest gains for the PL candidate were in Rio Grande do Sul (+6.75 pp), Tocantins (+6.44 pp), Ceará (+5.90 pp), Mato Grosso do Sul (+5.90 pp) and Mato Grosso (+5.30 pp). In Goiás, Lula's drop was not accompanied by a rise for the PL candidate (+1.45 pp); the reason is in finding 6.

The votes tell a different story, because the states differ greatly in size. In São Paulo, Lula has 984,619 fewer votes than in 2022 and the PL candidate, 682,034 more. In Minas Gerais, it is 613,635 fewer for Lula and 538,284 more for the PL candidate; in Rio Grande do Sul, 511,913 fewer and 328,760 more. The PL candidate's biggest vote gains are not all in states where he leads: in Bahia, where Lula received 66.17% of the valid votes, the PL candidate has 394,986 more votes than in 2022; in Ceará, 379,023 more.

The Northeast. Lula's share fell in all nine states and the PL candidate's rose in all nine, but Lula kept the lead in every one, from 54.73% in Alagoas to 70.99% in Piauí. The drop was smallest in Sergipe (−1.08 pp) and Alagoas (−1.77 pp) and largest in Maranhão (−4.85 pp) and Bahia (−3.55 pp). Abstention, for its part, fell in seven of the region's nine states (the exceptions were Paraíba and Rio Grande do Norte).

States that switched sides. Minas Gerais went from a Lula lead of 4.7 pp in 2022 to a PL candidate lead of 4.9 pp in 2026. Tocantins went from 6.4 pp in Lula's favor to 7.0 pp in the PL candidate's favor. These two switches account for the difference between 14 and 12 federative units for Lula.

Municipal maps. In 5,468 of the 5,570 municipalities, Lula's share fell; in 102 (23 of them in Pará, 18 in Bahia and 9 in Sergipe), it rose. The PL candidate's share rose in 5,479 and fell or held steady in 91. The biggest changes appear in small municipalities. The biggest drop in Lula's share was in Guarani de Goiás (GO, 4,085 inhabitants), at −24.5 pp (the PL candidate, −1.4 pp). Next come Herveiras (RS, 2,565 inhabitants), at −24.1 pp for Lula and +21.6 pp for the PL candidate, and Passa Sete (RS, 3,983 inhabitants), at −22.4 pp and +22.1 pp. All are municipalities with fewer than 5,000 inhabitants, where a few dozen votes move the percentage a lot.

The state capitals. Lula's share rose in 11 of the 27 capitals and fell in 16. The biggest drops: Goiânia (−4.86 pp), Palmas (−4.61 pp), Porto Velho (−2.47 pp), Cuiabá (−2.41 pp) and Campo Grande (−2.19 pp). The biggest gains: Macapá (+2.41 pp), Belém (+2.26 pp), Aracaju (+1.50 pp), Brasília (+1.26 pp) and Recife (+1.25 pp). The PL candidate's share fell in four capitals: Goiânia (−1.41 pp), Recife (−0.85 pp), Brasília (−0.34 pp) and Rio Branco (−0.27 pp). In São Paulo, Lula's share fell 0.96 pp (from 47.54% to 46.58%) and the PL candidate's rose 4.14 pp. In Salvador, Lula took 65.26% (−1.73 pp).

In terms of votes, the city of São Paulo accounts for Lula's biggest loss (224,163 fewer votes than in 2022), followed by Salvador (78,370), Belo Horizonte (45,965), Rio de Janeiro (43,968) and Goiânia (39,626).

Municipalities that break from the state pattern. To find municipalities that diverge from what happened in the rest of the state, we calculated the change in Lula's share in each municipality minus the (weighted) median for the state and listed those farthest away. Among those that fell more than the state: Senador José Porfírio (PA, −16.5 pp), Humaitá (AM, −13.1), Eirunepé (AM, −13.1), Jutaí (AM, −12.8), Colina (SP, −13.2), Anapu (PA), Rurópolis (PA), Tailândia (PA), Balsas (MA, −13.3) and Papanduva (SC, −12.6). Among those that did better than the state: Sobral (CE, +4.2 pp), Limoeiro do Ajuru (PA, +4.2), Salvaterra (PA, +3.2), Anajás (PA, +2.7), Quinjingue (BA, +1.6), Belém (PA, +2.3) and Baixa Grande (BA, +1.5). We do not know the cause in any of these cases. Before drawing any conclusion, it is worth checking whether there is a local candidacy, a change of electoral zone or another factor that the TSE data do not show. The case of Sobral, which has a measurable explanation, is in finding 6.

By municipality size. The smaller the municipality, the bigger the change. In municipalities of up to 5,000 inhabitants (1,324), Lula's share went from 54.55% to 47.38% (−7.17 pp) and the PL candidate's from 39.28% to 46.06% (+6.77 pp). In the 41 municipalities with more than 500,000 inhabitants, Lula went from 45.03% to 43.98% (−1.06 pp) and the PL candidate from 44.01% to 46.41% (+2.40 pp).

Six findings outside the norm

Each finding includes what we tested to find out whether it is an artifact of the data.

Finding 1. The change was almost national, but two thirds of it is local. Lula's share fell in 5,468 of the 5,570 municipalities (98.2%) and the PL candidate's rose in 5,479 (98.4%). In 5,413 municipalities both movements happened at the same time. The average change in Lula's share across municipalities, weighted by 2026 valid votes, is −3.46 pp, and the weighted standard deviation, 3.04 pp. Only 34.7% of the variation between municipalities is explained by the state they are in; the rest, 65.3%, is variation between municipalities of the same state. Test: repeating with only the 2,088 municipalities with more than 10,000 valid votes, the drop holds for 2,024 of them (97%) and the part explained by the state is 33.5%. The pattern does not come from tiny municipalities, where a few dozen votes change the percentage.

Finding 2. The smaller the city, the bigger the change, and this holds within each state. In the regression of the change in Lula's share on the logarithm (base 10) of population, each tenfold increase in population comes with a smaller drop, of about 1.9 pp, already discounting the state. Artifact tests: (a) with state fixed effects, the coefficient is 1.92 (without controls, 2.09); (b) also controlling for Lula's share in 2022, to separate it from regression to the mean, it is 1.82; (c) with only municipalities with more than 10,000 valid votes, 1.93; (d) without the five largest federative units (SP, MG, RJ, BA, RS), 2.34; (e) in a random split of the set into two halves, 1.99 and 2.22; (f) in each of the 26 states with at least 10 municipalities, the sign is the same (positive, from 0.26 in Piauí to 3.04 in Santa Catarina). This is correlation: the data do not say why smaller cities changed more.

Finding 3. Abstention barely explains the change in votes, and the national correlation is misleading. We have already shown that, in the decomposition of votes, abstention weighs little. At the municipal level, the change in abstention has a correlation of −0.24 with the change in Lula's share (more abstention, less Lula). But within each state the correlation is +0.14, the opposite sign. It is a reversal caused by mixing states. Across the 27 federative units, the correlation between the change in abstention and Lula's share in 2026 is −0.30: in the states where Lula has a larger share, abstention fell more or rose less. This (guess) creates an aggregate negative correlation that does not exist within each state. Test: with the 4,968 municipalities whose electorate changed by less than 10% between the two elections (discarding 602 with a larger change, which could be an effect of voter re-registration), the within-state correlation is +0.18, in the same direction. Splitting the municipalities into four groups by the change in abstention relative to their own state's average, Lula's share was 0.98 pp below the state average in the group where abstention fell most and 0.57 pp above it in the group where it rose most; for the PL candidate, the deviations were +0.75 and −0.37 pp. The data do not say who abstained or why. The deviations are small compared with the total change of 3 to 5 pp.

Finding 4. 714 municipalities switched winners, all in the same direction. From Lula in 2022 to the PL candidate in 2026: 714. None made the reverse trip. Together they account for 8.19 million valid votes in 2026 and are concentrated in Minas Gerais (178), Rio Grande do Sul (134), Paraná (91), São Paulo (90), Goiás (69), Tocantins (37) and Santa Catarina (27). The median 2022 margin in these municipalities was 7.0 pp, versus 33.1 pp in the municipalities that did not switch: these were switches in races that were already tight. Test: we applied to the 2022 margins a uniform shift, the same for all, of 7.4 pp (the weighted average shift between the two years). This predicts 394 switches, of which 385 coincide with real switches. The uniform shift gets almost all the switches it predicts right, but leaves out about 330 of the 714: in those places the shift was larger than average. Restricting to municipalities with a 2022 margin of at least 1 pp and at least 5,000 votes, 3,405 municipalities remain, of which 292 switched. In 2026, 426 municipalities ended with a margin of up to 5 pp (83 with up to 1 pp).

Finding 5. Within a city, neighborhoods change almost as much as cities within a state. We matched 6,082 electoral zones (local voting districts) between 2022 and 2026 and kept the 3,988 whose electorate was between 0.8 and 1.25 times its 2022 size and that have at least 5,000 valid votes in 2026. In the 73 municipalities with 3 or more of these zones, the median difference between the zone that changed most and the one that changed least is 2.24 pp (90% of municipalities are below 4.85 pp). The standard deviation of the change across zones of the same municipality is 1.48 pp; across municipalities of the same state (with more than 10,000 valid votes), 2.25 pp. In São Paulo, the change in Lula's share by zone ranges from −5.0 to +4.9 pp; in Brasília, from −1.3 to +5.7; in Rio de Janeiro, from −3.2 to +5.0; in Goiânia, all comparable zones fell, from −7.3 to −1.7. Test: we excluded zones whose electorate fell outside that range, so that the change is not an effect of zone redrawing. The numbers by zone are in zonas_estaveis_2022_2026.csv.

Finding 6. Third candidates explain the biggest drop, and a 2022 candidate explains a "gain" for Lula. Candidates who are neither Lula nor the PL candidate totaled 7.81% in 2026 and 8.37% in 2022: little changed in total, but the distribution is not uniform. Ronaldo Caiado received 2.18% nationally and 12.33% in Goiás (1.85% outside Goiás); in Goiás, candidates outside the two main ones totaled 15.3% of the valid votes (8.3% in 2022). It is in that state that Lula has his biggest drop (−8.46 pp), while the PL candidate rose only 1.45 pp. Augusto Cury (2.89% nationally) received 4.10% in Amapá, 3.59% in São Paulo and 3.57% in Amazonas; Renan Santos (2.24%), 3.23% in São Paulo, 2.95% in Santa Catarina and 2.69% in Paraná.

The other side: in Sobral (CE, 203,000 inhabitants), Lula's share rose 4.2 pp, the municipality with the largest positive deviation from Ceará. In 2022, Ciro Gomes received 18.46% of the valid votes in Sobral (6.80% in Ceará and 3.04% nationally) and did not run in 2026. In Fortaleza, he received 8.66% in 2022; Lula's share there rose 0.28 pp and the PL candidate's 4.17 pp. Test: the hypothesis that this is the effect of a local 2022 candidate is consistent with the numbers (guess: Ciro's votes in 2022 were redistributed between the two), but we cannot measure where they went. When only the two main candidates are compared, the 2022 and 2026 data are not equivalent in states where there is a strong third candidate in one of the years.

What news outlets and institutes say

We separate what is data (ours, with a source), what is someone else's interpretation (with a link) and what is our own reading.

Public data from another source checked against our numbers. Congresso em Foco reports that Flávio Bolsonaro won in 15 federative units (14 states and the Federal District) and Lula in 12 (the nine states of the Northeast plus Pará, Amazonas and Amapá). Our calculations, from the TSE file, give the same list (Congresso em Foco). Note: the article's headline and its closing have different counts (14 to 13 in the closing), but the list of states matches 15 to 12.

The outlet's interpretation. Congresso em Foco describes the result as a "divided Brazil" geographically too, says the Northeast is the only regional bloc won entirely by Lula, that the PL candidate's map was especially homogeneous in the South, Southeast and Center-West, and that the North is the most divided region. These adjectives are the outlet's. Our data show the following about the North: Lula received the most votes in 3 of 7 states (Pará, Amazonas and Amapá), and the margins are between 0.05 and 5.4 pp in the three.

Interpretations of abstention. They are in section 2, with authorship, link and an indication of what is opinion, study or the outlet's own analysis. Our reading is only what the data support: abstention rose 0.13 pp, stayed close to the 2018 and 2022 level, and does not explain the change in votes between the two candidates.

Voting intention polls for the runoff (data only, no reading of our own). We reproduce only the technical data of two polls registered with the TSE, as released. A poll is neither a forecast nor a result, and we do not comment on the numbers.

In both, the percentages are of all respondents, and the remainder is blank, spoiled and undecided (Datafolha: 6% blank and spoiled and 2% who did not know; Quaest: 9% blank, spoiled or not planning to vote and 2% undecided). Quaest did not report valid votes. Sources: SBT News and Correio da Manhã for Datafolha; Gazeta do Povo for Quaest. The original numbers are on the institutes' websites and in the TSE's poll registration system; check there before citing.