Did Brazil's runoff extension shut out Bolsonaro voters?

What happened on Sunday afternoon

Both campaigns asked for more time. Lula’s cited reduced bus service in cities such as São Paulo and Salvador, which tend to favor him, while Flávio Bolsonaro’s pointed to heavy rain and flooding in the South, where he is favored, and asked for an extension in 17 municipalities of São Paulo, Paraná and Santa Catarina (PBS, Anadolu).

At 17h01, after the legal closing time, Justice Cristiano Zanin ordered voting extended to 19h in every section of the country. Because many machines had already closed and printed their results, he then clarified that the order covered only sections still open, and the electoral court relayed that guidance to the regional courts at 17h49. By then the courts of the Federal District and Mato Grosso do Sul had said they would not extend at all (Conjur), and São Paulo’s regional court reported that 80% of its sections had closed at 17h (Poder360). Since Lula was declared the winner, Flávio has said he will ask the courts to annul the result, which makes it worth asking what the machines themselves recorded.

Where the extension actually happened

Whether a section kept voting depended on whether it was still open when the order arrived, and on how its state’s electoral court chose to act. Most of the South and Center-West, where Flávio is strongest, effectively had no extension, while Rio de Janeiro, the Northeast capitals and the city of São Paulo account for most of it.

Did Bolsonaro-leaning sections close more often?

Nationally, a little, since 1.23% of sections where Flávio led in the first round stayed open against 1.78% of those where Lula led. That gap, however, comes entirely from which states and cities extended. When sections are compared only with others in the same municipality or electoral zone, a section’s first-round lean has no measurable bearing on whether it stayed open.

Inside the cities that extended

Within a single city the pattern runs in both directions. In Rio de Janeiro, Flávio-led sections were more likely than Lula-led ones to stay open, while in Salvador and Niterói the reverse held, which is what one would expect from sections that happened to have lines at five o’clock rather than from any political sorting.

How much it could have mattered

Almost every urna closed within minutes of 17h, before the order reached the polling places, and none closed early. To bound the effect of the uneven extension, suppose every urna that closed on time had stayed open and drawn late voters at the rate the open ones did. Even if every one of those votes had gone to Flávio, the total would cover a small fraction of the margin, and splitting them by each section’s own lean leaves almost nothing.

How this was checked

Closing times come from the log of every urna (logd.dat), which TSE publishes on its results site. Two independent public crawls, urnas-tardias and logsdasurnas, agree on 99.9% of urnas and reproduce the press count of about 34,000 votes after 17h (CNN Brasil). Times are converted to Brasília time using each municipality’s usual opening hour, and an urna counts as extended if it closed after 17h45, well past the few minutes normally needed to clear a line.

Political lean is each section’s own first-round vote on 4 October, taken from the TSE boletins de urna and measured before anything happened on 25 October. All 497,882 domestic urnas matched one to one, and overseas voting is excluded. The like-for-like comparison is a linear probability model with fixed effects for state, municipality or electoral zone, so each section is compared only with others that faced the same regional court and the same local conditions.

What the data cannot show is how many people arrived at a section after it closed, because the logs record votes rather than voters who were turned away, so the ceiling chart bounds that number instead. The closures themselves were lawful, since polls close at 17h and the order kept open only what was still open.