Election polls: how are they done and how reliable are they?

Opinion polls gain importance on eve elections. Polls on the population’s intentions to vote for candidates running for political office win the press by their numbers, which serves to guide the actions of campaigners in search of better results.

In the midst of so many issues released each week, it’s easy to get confused, and some voters take the easy route: distrust. After all, if each survey has a different result, how will you know which one is closer to the truth?

“Election polls measure intent to vote at the time they’re conducted. Even if they are well done, there are several factors that could cause them to miss out on survey results,” says statistician Neale Ahmed El-Dash, developer of Polling Data, a research aggregation site.

“If there are too many undecideds in the poll, on election day these voters will go somewhere – they may or may not vote, but they will not continue to rank as undecided. Some candidates may be in the poll but not on Election Day. If the poll is conducted long before the election, the chances of the result being right are also reduced, as events during the campaign can change voting intentions,” he explains in an interview with El-Dash. TecMundo Made via WhatsApp.

More than a dozen institutes in Brazil are conducting election polls. Outside of the election period, these companies conduct different types of opinion polls commissioned by brands, governments, and other organizations.

While there are slight differences in the way they conduct an election poll (some done in person and others over the phone, for example), they should all agree on one aspect: because it is not possible to interview more than 156 people. If there are millions of Brazilian voters, the sample (group to be interviewed) should reflect all national voters.

“The people to be interviewed are selected to fulfill the quotas defined in the sample. “Quotas are proportional to variables of gender, age, education and employment status,” Márcia Cavallari Nunes, CEO of Ipec (Intelligence in Research and Consulting) says in an interview. TecMundo via email.

Ipec was founded in 2021 by former Ibope executives following the closure of Ibope. Because Ibope is well known to the Brazilian public, Ipec is often referred to in the press as the old Ibope.

But how does the institute define, for example, how many women should be heard or how many people over 30 should be interviewed? Nunes explains that the profile of the sample is defined based on official data from the Pnad (National Survey by Household Sample from IBGE), which is conducted annually. Data from TSE (Supreme Election Court) are also taken into account.

“It’s not the sample size that matters, but how the people are selected,” says El-Dash.

The statistician uses a culinary example to illustrate the point. Imagine preparing a lasagna with several layers of filling and dough. To know the taste of the preparation, it is necessary to cut a slice, albeit very small, that covers all layers of the plate. Otherwise – if you only get one layer – you won’t be able to know if the lasagna is good, you’ll only have a partial view.

For this reason, the surveys made by social network users are far from reflecting the opinion of the country’s population. In our social relationships, we tend to get closer to those who hold similar views (at church, school, etc.). These are so-called “bubbles” that make us think the world around us is less diverse than it really is.

“Who should control who will respond to the survey, who will be in the sample, it is the statistician who designed the sample, not the respondent,” El-Dash says. Therefore, it is not right to insist that pollsters have their voices heard in an election poll; they have the education and know how to select respondents so that the sample can better reflect the views of the population. Voting intent polls are moderated and must be registered with electoral courts.

In person or by phone?

For al-Dash, both methods have their challenges and benefits. “It is not an obvious thing to say which method is better. But the evidence I’ve found suggests that depending on the subject and interest, phone surveys can be a little better,” says the statistician.

“Online surveys were the gold standard [considerado o melhor método] In the past and knowing the opinion of Brazilians it was very important, but the trend for the future is to rely on telephone surveys. It’s faster, cheaper, and at least as efficient as face-to-face meetings,” adds El-Dash.

In recent years, poll aggregators have emerged to help voters navigate data. On these platforms it is possible to have a broader view combining the results of various surveys. In addition to the Survey Data website, similar platforms (with different methods) have been developed by the newspaper. State of Sao Paulo (Estadao)and through websites Power 360 and CNN Brazil.

For al-Dash, the growing distrust of election polls in recent years is more a reflection of personal opinions than failures in the polls.

“No matter how much they question the election polls, people actually criticize a result that is not what they want, not methodology or performance. This is not a sign of any scientific thought,” says El-Dash.

“People criticize an outcome that isn’t what they want, not methodology or performance.”

In extreme cases, the population turned violently towards the researchers of the institutes. On September 20, a researcher from the DataFolha institute, one of the country’s most important institutes, was kicked and punched in Ariranha (a city in the interior of São Paulo) by a supporter of President Jair Bolsonaro (PL). re-election.

Márcia Nunes of Ipec confirms that attacks on researchers in this field are increasing. “We have had cases of verbal and even physical aggression such as pushing. “We didn’t have these kinds of seizures before, they were rarer, they happen more often today,” says the Institute’s CEO.

Opinion polls are indispensable for the elaboration of public policy in a democracy, and even more so in various countries such as Brazil. Although questionable, the population should make their objections within the law, demanding greater transparency (where applicable) and methodological rigor so that the results are most reliable.