The frenzy around election polls increased in 2018. Unqualified accusations of manipulated results and of poll results being used for market speculation raised a red flag on the topic.
Last week, messages between financial traders discussing politics were made public. In them, some traders - supporters of Jair Bolsonaro (PSL) - said that people shouldn't guide their investments on election polls because they were biased.
According to the message exchanges, the polls oversize the size of the northeastern population who earns up to two times the monthly minimum wage; they bring down the number of respondents who claim to be Catholic and evangelic and interview more people who declare to be left wing. These sampling errors would be beneficial to candidates like Fernando Haddad (PT) and Ciro Gomes (PDT).
Datafolha's statisticians rebutted the criticism point by point, explaining that they use concepts and techniques based on sampling theory: "The samples are representative of the studied population and selected with statistical criteria, using official government data as sources," they said.
Polls are one of multiple influencing factors on a voter's voting choices. They talk to relatives and friends; they assess their current living conditions and project the future ahead. They watch the debates and TV ads. They have ample access to campaigns news and react to what the candidates say.
One new thing in the current election is that we have more polls being made, many commissioned by big financial institutions, which are privy to the results before they are disclosed to the public. This embeds the risk of certain groups influencing the markets since they have access to the poll data beforehand.
Setting aside regulatory issues, it's a fact that with so many polls, there as just as many methodologies.
There are three ways to approach a respondent: in a high foot traffic area, at their house or by phone. Polls commissioned by financial institutions are usually made by phone, thus showing relevant discrepancies when compared to the other two.
Datafolha chooses the high-traffic approach. Director Mauro Paulino says that surveys made in person, either in the street or at home, are better, and usually show similar results when the margin of error is considered.
In his view, telephone polls have limitations. "They tend to privilege candidates who appeal to elites. People of low income often can't stop working to answer a survey, even when they have a cell phone," he explained.
After so many elections and a certain poll fatigue, readers are now understanding surveys better and demand more precision in the data disclosed.
Last week, Folha published two corrections originated on technical mistakes. One of the mistakes was in the Sept. 15th headline: "Bolsonaro Goes To 26% And Haddad Is Tied With Ciro And Alckmin". Fernando Haddad (PT) and Ciro Gomes (PDT) had the same 13%, but Geraldo Alckmin had 9%. Because the tie belonged in the limit of the margin of error, it was wrong to say that Alckmin was tied with the other two.
Regarding second-round possibilities, Folha published that Marina Silva (Rede) and Alckmin would be tied with Jair Bolsonaro (PSL), but in fact, they were leading.
Managing editor Vinicius Mota reminded me that it has been agreed to not use consider a tie in news articles when there is a difference between candidates of twice as much as the margin of error. In this case, the likelihood of a tie is minimal.
"To issue corrections, especially in relevant news pieces like this one, is never a pleasant thing to do," he said, highlighting the fact that the error was corrected as soon as it was spotted.
In such a cutthroat campaign, these are mistakes that even when promptly corrected, damage the papers and the institute's reputations. Readers will instantly come up with conspiracy theories and bias accusations.
This kind of distorted interpretations, as well as fake opinion polls, are in the heart of fake news, and useful to the dirty game played by the campaigns and the financial markets.
Translated by NATASHA MADOV