O que precisa mudar nas pesquisas
Todos os institutos de pesquisa apontavam Flávio Bolsonaro como o vencedor da eleição de domingo. Os prediction markets eram ainda mais otimistas com o senador: plataformas como Polymarket e Kalshi chegaram a atribuir probabilidades superiores a 80% à sua vitória.
Recent Brazilian elections saw the prediction markets and polling institutions incorrectly predict Flávio Bolsonaro as the winner. These institutions overestimated his chances of victory to over 80%. However, President Lula was reelected with 51.79% of valid votes, a margin of 4.4 million votes. This outcome reignited a discussion that has arisen in past elections, specifically about the accuracy of polling methods.
Carlos Augusto Montenegro, former president of Ibope, suggested that the problem begins with the speed at which voters change their opinions. He noted that a campaign event does not end with a television debate, but rather continues on social media and conversations among friends and family, influencing undecided voters. Montenegro believes this volatility helps explain why all polling institutions were surprised in 2026, even with different methodologies.
Clifford Young, president of Ipsos research, believes Lula's victory was more related to voter mobilization and a decrease in abstention, with 3 million more voters compared to the first turn. He acknowledges that the industry likely did not capture this movement and proposes improving likely voter models, which aim to identify not just who a person intends to vote for, but the probability of their actual turnout.
Renato Meirelles of Instituto Locomotiva believes the difficulty goes beyond turnout and that the decision-making process of voters has changed. He argues traditional questions about voting intent need to be accompanied by information on satisfaction with the candidate, intensity of preference, and willingness to change votes. Meirelles suggests combining qualitative and quantitative research, as well as tracking the same individuals throughout the campaign to identify shifts in preferences and what causes them.
Montenegro also supports changes, suggesting a combination of different methodologies - in-person, phone, and internet interviews - instead of relying on a single method of data collection. He believes that, given the volatility of the electorate, a face-to-face poll taken on election day could once again make sense.
Written by urgent.news from Brazil Journal's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
