There were a lot of eyes on Brazilian pollsters today. More than usual in a country with more than 200 million souls and one of the most powerful economic engines in the world. The election between Lula and Bolsonaro could hardly put more things at stake: the stability of the largest democracy in Latin America; the political control of key resource sources (including as a “resource” the capture of CO2 and the emission of oxygen from the Amazonian territories); the definitive battle between left and right (although it is neither the first nor will it be the last, the profile of this one was especially clear); and, of course, the future of Brazilian citizenship. All those questions boiled down to one: who was going to win and by how much. The consensus of the polls was correct in the first and remarkably close in the second: Lula would win, and he would do so narrowly.
During the last week, the average of polls put the former president ahead between 1.4 and 2.6 points. The final result would end up in that narrow band: 1.8. This is truly the toughest test of accuracy that can be asked of a tight second-round election.
Especially after what happened last October 2nd. So the polls were right about everything (who would be first, who would be second, if there would be a second round) except for one thing: the level of Bolsonaro’s vote, which they significantly underestimated. The current president took advantage of the circumstance to finish off his main battle against all electoral, state or civil society institutions. This error gave him a perfect argument to continue feeding his thesis of “I will win, if nothing strange happens”. In this case, the failure of the polls counted in that “rare” drawer according to Bolsonaro, despite the fact that errors of this magnitude are common in all countries, moments and ideological directions imaginable, almost always due to methodological biases. But the fact is that this line of speech put an even more intense target on the polls ahead of the second round.
In fact, new ones entered the group, and three of them anticipated a Bolsonaro victory until the last moment. Despite this fundamental failure, they have not been the ones that have obtained the worst aggregate absolute error (the result of adding the errors, positive or negative, that the last published poll of each house had on both candidates).
Join EL PAÍS to follow all the news and read without limits.
subscribe
This suggests that, indeed, the most established pollsters continue to have problems in calibrating Bolsonaro well. It is worth mentioning that most of them have maintained a high volume of probable voters assigned to the category of undecided until the last moment: more than 6%, something like the equivalent sum of those who chose a third candidate last 2 of october. As a result, both candidates were actually underestimated on average, although the winner was less so than the loser.
In fact, if the average error is calculated with undecided voters and the average error without them (as if they did not count for the calculation of the voter base), the error is notably reduced until it almost disappears. The average of polls without undecided almost predicted the final result, with a few tenths of excess for Lula and defect for Bolsonaro.
All of this suggests that the Brazilian election has been yet another example of testing polling methods in an extremely polarized environment. If in this case it has been overcome, it is probably due to a combination of factors: being able to publish until almost the day of the vote (which better captures last-minute decisions), having a plural staff of pollsters (larger in the second return than in the first), and perhaps (but only each house knows this) adjustments made between October 2 and 30. But that high percentage of unassigned likely voters still hung like a sword over the forecasts. One that could have fallen from the other side, becoming available to be wielded once again by those who feel they have something to gain from data damage.
Subscribe here to newsletter of EL PAÍS America and receive all the informative keys of the current affairs of the region
Subscribe to continue reading
read without limits
#Polls #regain #credibility #Brazil