Specialist analyzes online research and its limitations

According to the specialist interviewed by Estadão, the hypothesis is that online surveys tend to have a more restricted coverage of the population, as only people with internet access participate in them. The expert explains that in the case of Atlas, the questionnaire is accessed through ads on Google and social networks, which further limits the sample, as it depends on who sees and clicks on the ad. Another issue is that people who tend to respond to the Atlas survey are more politically engaged, while Quaest is less susceptible to this bias, Nishimura says.

The director further explains that there are specific difficulties for each approach. “The household survey, like Quaest’s, theoretically can reach anyone, but in Brazil there are difficulties in recruiting certain respondents, especially in areas with gated communities and buildings where interviewers cannot enter. Atlas, on the other hand, faces more obstacles with low-income or elderly voters who are not as connected to the internet,” the director remarks.

Nishimura highlights that both institutes adjust their samples to deal with these limitations: for example, Atlas tends to give more weight to lower-income voters, compensating for the fact that they tend to capture less of this profile in their surveys. Quaest, on the other hand, tends to compensate for the underrepresentation of voters with higher purchasing power.

The ‘shy voter’ hypothesis may impact results

Nishimura also mentions the “shy voter” hypothesis, popularized in the 2016 United States elections, but still lacks scientific evidence. “The idea behind this hypothesis is that some voters, in face-to-face interviews, may feel embarrassed to declare their vote for certain candidates, while in online surveys, without the presence of an interviewer, they tend to be more honest. However, this hypothesis has never been supported by solid systematic evidence,” Nishimura emphasizes.

Regarding the impact of the shy vote in the São Paulo scenario, he suggests that this difference could affect candidates like Pablo Marçal, more associated with an extremist electorate.