Brazil: the polls missed in both rounds
They underestimated Flávio in round one and missed Lula's win in the runoff.

Brazilian pollsters had a tough election, with misses in opposite directions in each round. In the first round on October 4, the main firms got Lula’s share right, around 45%, but underestimated Flávio Bolsonaro, who finished with 47.06% in that night’s partial count against the president’s 45.12%.
Metrópoles compared the final numbers. Quaest, with 3,702 interviews on October 2-3, had Lula at 46% and Flávio at 45%. AtlasIntel, with 4,945 interviews from September 27 to October 2, had Lula at 47% and Flávio at 44.1% of valid votes. Datafolha nailed Lula’s 45% but underestimated Flávio by about five points. In short, Flávio ended two to five points above what the polls showed.
The runoff went the other way. On the eve of October 25, Datafolha and Quaest had Flávio at 51% and Lula at 49% of valid votes, and AtlasIntel, with 4,990 interviews from October 18 to 23 and a one-point margin of error, showed 50.8% to 49.2%. Lula won with about 51.8%: the polls missed the comeback and, at best, showed a statistical tie with Flávio ahead.
Explanations point to late shifts. For the first round, Metrópoles attributes the gap to weekend strategic voting: supporters of other right-wing candidates such as Renan Santos, Ronaldo Caiado and Romeu Zema moved to Flávio. In the runoff, abstention fell from 21.08% to 19.21%, and hundreds of fan groups campaigning for Lula set out to reach those who had not voted or had spoiled their ballots.
For pollsters and strategists, the case holds technical lessons. In polarized, close elections the difference between winning and losing fits within the margin of error, and shifts in the last 48 hours, such as strategic voting or changes in turnout, fall outside published measurements. Likely-voter models and the way undecided voters are allocated matter more than sample size.
The thing to watch is the methodological review the firms themselves carry out after two consecutive misses in opposite directions. For campaigns, the practical takeaway is to measure until the last day with in-house tracking, and not to treat the public average as a diagnosis.
In polarized races, late movement and turnout matter more than the average: track to the last day.
Summary written by the FOCUS editors based on the original story. Rights to the original content belong to its publisher.
FOCUS courtesy space


