a climate of mistrust fostered by Jair Bolsonaro A statistician at the University of Michigan said the poll may have been skewed. The current president won more votes than expected.
Results of this Sunday’s election Brazil Investigators were so wrong in their predictions of President Jair Bolsonaro’s turnout that they surprised those who had relied on the election results.
But for Raphael Nishimura, Director of Sampling at the Center for Research at the University of Michigan (USA) and a member of the American Association for Public Opinion Polling (Aapor), it doesn’t make sense to treat election results as an oracle.
Deutsche Welle: Were the polls wrong?
Raphael Nishimura: Pre-election voting begins a day or two in advance [de la votación]They always portray the voters of the moment. If there is a change in votes between the completion of voting and the election, either because voters decided to vote at the last minute, or because some people cast their votes and change their votes based on them Obviously, no investigation will take place. I’ll take this last minute change.
This is the case in all elections.
So what happened?
One possible explanation could be a last-minute move in voter opinion, but the move was not caught because the polls ended before the move.
Another hypothesis is the possibility of no response. That is, if someone does not want to respond to surveys. This is highly dependent on the respondent’s opinion of what is being asked.Bolsonaro’s vote may have been underestimated [debido a esto].
Presumably, these voters will have lower voter participation than those of other candidates. Due to the lack of data and empirical evidence on this point, it is a very difficult hypothesis to prove. It’s a very difficult bias to correct, even with statistical methods.
[Este fenómeno] What happened in the US in 2020 was due in part to Donald Trump’s own narrative that the polls were all bought out. Some of the voters we hear may have decided not to answer these polls. Because then, obviously, the polls won’t be able to capture the Trump vote very well.
It becomes a self-fulfilling prophecy. Bolsonaro has used a narrative of very similar questions to discredit the polls so that his most ardent voters may not want to answer the ballot.
Hypothesis. And consistent with this hypothesis is the fact that the vote for Lula is closer than most polls indicated.
If polls aren’t predictions, what are they for?
The simple fact of knowing the voter’s current scenario, their preferences, is already important information. We are not only dealing with predictions of the future, but also scenarios that have already happened and current scenarios. And if you run pre-election surveys over time, as many agencies do, you can observe trends.
Will the discrepancy between this poll and the election results have any impact on Round 2?
Hard to say. On the one hand, in the short term, I think it’s hard to change what pollsters are doing now. I expect that they will try to improve some aspects of their methodology for the second round.
It is this hypothesis of lack of response that worries me the most. Not only is it exacerbated by some voters’ perception that polls are biased, making it a self-fulfilling prophecy, but it’s also exacerbated by Bolsonaro’s own statements. Affirming that it has beaten research institutions, it makes these voters even less likely to respond to polls, which could exacerbate the problem.