The polls this year have a bias... What we are seeing in some polls in Brazil goes through a methodology called likely voters. Basically, the institutes try to estimate not only who the voter intends to vote for, but also the probability that this voter will actually show up at the polls. And here's an important point: Lula performs very strongly among lower-income voters, a segment that has historically been more prone to abstention. When this factor enters the model, the tendency is to reduce his advantage and create an appearance of greater closeness to Flávio. Does that make sense? More or less... The problem is that this type of methodology depends on something extremely difficult: predicting today who will effectively vote on Election Sunday. And 2026 may have a completely different dynamic from 2022. What we are seeing in terms of mobilization in the streets in favor of Lula, for example, is something that must be taken into account. Mobilization also affects the voter’s willingness to leave home to vote. Therefore, I have my doubts about to what extent this "likely voter" model will be able to accurately capture this election. The traditional method of voting intention has always worked in Brazil. In 2022, we had a very specific context, including PRF operations involving the transportation of voters that led the TSE itself to intervene during the second round. This needs to be remembered when someone tries to use that election as an absolute parameter to project abstention in 2026. So I would be very cautious with polls that try to turn voting intention into predictions of turnout. After all, polls measure intention. Who will or will not step outside on election day is a variable that is much more difficult to predict. (translated)
