The questioning of the work of research institutes, fired in Brazil after the first round of elections, with the right to attacks and maneuvers from the government base, is a phenomenon already well known in the USA.
The reason is the same: a mismatch between what the surveys indicate —which, moreover, do not aim to get the result right—and the numbers that come out of the polls.
In 2016, the data indicated an inevitable victory for Democrat Hillary Clinton, who ended up losing to Republican Donald Trump. In 2020, they said Joe Biden would win, but with a bigger advantage than he actually had — the institutes exaggerated the margin in the national popular vote by 3.9 percentage points for the Democrat and by 4.3 points in the state polls, for the Electoral College.
The matter returned to public debate with the arrival, in less than a month, of the midterm elections. To defend their reputation and improve their results, research institutes have been trying to adjust their estimates, in order to prevent the press and population from attacking them again.
The organizations want, for example, to avoid a disproportionate focus on high-educated populations — considered one of the possible lapses in 2016. They are also trying to hear more Republicans aligned with Trump. This share of the electorate, traditionally more averse to talking to researchers, may have been underestimated in 2020.
If it is enough, this will only become clear after the election. Analysts believe that the path is limited, if only because it is not entirely clear what exactly went wrong in the last elections or the best way to balance the discrepancies.
In the November 8 midterms, Americans vote to renew all 435 House seats and 35 of the 100 Senate seats. There are also contests for governor and other state positions.
Since World War II, the party that controls the White House has almost always lost ground in this election. Biden’s Democrats now have 220 seats in the House, a minimal eight-seat advantage, and 50 in the Senate, plus the mining vote given to Vice President Kamala Harris.
Given the president’s fragile popularity, the expectation was that Republicans would regain both houses, but some institutes have indicated another scenario. FiveThirtyEight, the leading poll aggregator, says Democrats have a 67% chance of retaining control of the Senate.
The imbroglio becomes clearer when looking at disputes individually. Surveys indicate a tight Senate race in Ohio, for example, but the state is now considered a bastion of Republicans. Something has changed there — or the polls aren’t quite right.
Mark Weaver, an Ohio-based election attorney, believes the institutes may be underestimating the Republican electorate. As in the past, he says, polls hear far too many people with a college degree, but the proportion of the population with that level of education is 29% in the state.
This matters because researchers believe that highly educated people tend to vote Democrat. “If we want to know what people think about politics, we have to ensure that the data reflect what we already know about the population,” he says.
Another problem, according to Weaver, is that Republican voters — particularly those who voted for Trump and allies — have been increasingly resistant to participating in these polls. “They’ve learned to shut their mouths when someone starts discussing politics,” she says, recalling that part of the press and the government itself describe them as irrational. “People lie or avoid talking to researchers. This is a systemic issue, which cannot be resolved by changing the method.”
A similar phenomenon is raised in Brazil with supporters of President Jair Bolsonaro (PL) — not by chance, an ally of Trump.
To resolve issues in this field, several institutes announced revisions to their methods. The Pew Research Center, for example, said in 2021 that it was changing the composition of research groups. Specific sectors, such as highly educated Democrats, were reduced. At the same time, the institute looked for ways to listen to underrepresented groups, such as low-income people.
Analysts also suggest other strategies, such as giving different weight to Democratic and Republican voters, to correct eventual distortions. “When you expand such a field, however, the margin of error grows,” says Weaver. “It’s like when you xerox a photo and then enlarge it.”
To talk to the more skittish voters, one solution is to listen to them over the internet or by mail rather than by phone, to assuage their distrust. The understanding is that, speaking with a researcher, the voter may feel pressured to give a socially accepted answer.
The Marist Institute, which surveys American outlets such as NPR and PBS, has started approaching respondents via text message in recent months. “It has become more expensive and more difficult to do polls of intention to vote,” says Lee Miringoff, director of the entity.
Patrick Murray, director of a polling center at Monmouth University, points to another challenge: polls tend to affect voters. In the US, unlike Brazil, voting is optional. In 2016, when it seemed obvious that Hillary would defeat Trump, some Democrats may have stopped going to vote — and she lost. “This is especially true when you think the outcome is already decided.”
He defends the institutes, anyway. Popular dissatisfaction, he says, also has to do with a misconception about what researchers do. “Polls are not unambiguous predictions of outcome, but a picture of how you would vote in that particular circumstance,” Murray explains. “The problem is not whether research is reliable or not, but our expectations of what it might say.”
With a wealth of experience honed over 4+ years in journalism, I bring a seasoned voice to the world of news. Currently, I work as a freelance writer and editor, always seeking new opportunities to tell compelling stories in the field of world news.
