Election polls: between method, suspicion and evidence

Surveys are among the best tools available for methodically understanding what a society thinks at a given moment. Destroying their credibility weakens democracy, because democracy, in addition to votes, needs reliable ways to listen to, measure, and understand citizens before elections.

In election season, few things ignite passion like a poll. When the results favor one's own candidate, they are celebrated as proof of reality; when they are detrimental, they are dismissed as manipulation. However, before polls become objects of permanent suspicion, it's worth remembering something basic: probabilistic household surveys are not a capricious invention of politics. They are a fundamental tool of the social sciences and public administration.

The same type of statistical method used to measure voting intentions is also employed to measure unemployment, poverty, living conditions, nutrition, access to services, and many other social realities. Generally speaking, the procedure consists of randomly selecting municipalities, with a probability proportional to their population; then, within those municipalities, selecting neighborhoods, city blocks, or households; and finally, interviewing people following predefined rules.

This method is not perfect—no human instrument is—but it is robust, proven, and widely accepted. That's why it's strange, and sometimes downright obscurantist, that only when it comes to electoral opinion is there an attempt to discredit a methodology that the State, academia, and international organizations use every day to understand social reality.

Surveys are more closely monitored than many other measurements.

Election polls also have a unique characteristic: they are probably the most closely monitored social surveys. They are regulated by law, must meet specific requirements, and are reviewed by technical bodies. In Colombia, many polling firms also hold quality certifications, such as ISO 20252, and are audited by external entities that verify compliance with protocols.

This does not mean that a poll cannot be wrong. Of course, it can. All statistical measurements have margins of error, design limitations, and risks of interpretation. But it is one thing to acknowledge these limitations and quite another to suggest, without evidence, that the firms deliberately introduce malice to favor or harm a candidate.

That accusation is serious. And precisely because it is serious, it demands proof. It is not enough for someone to dislike the result. It is not enough for a campaign to feel that it is doing better "on the ground." It is not enough for a political leader to believe that their internal enthusiasm equates to a national majority. Accusing a polling firm of manipulation without evidence is not technical criticism: it's public irresponsibility.

The difficulty of measuring a shifting public opinion.

Election polls differ significantly from other social surveys: they measure an extraordinarily volatile phenomenon. Poverty, unemployment, and demographic composition don't usually change dramatically in a matter of days. Public opinion, however, can.

A campaign, a debate, an alliance, a scandal, fake news, or a collective emotion can shift the landscape in a very short time. Politics is sometimes like a horse race: trends can be seen for much of the race, but the final stretch can be decisive.

That's why polling firms often oppose excessively long blackout periods. They know that in the end, their polls will be compared to the election results, but often the law prevents them from publishing the image closest to the finish line. It's like asking a photographer to be responsible for the finish of a race, but prohibiting them from capturing the final hundred meters.

In short: election polls use proven methods, are subject to audits, and, moreover, face a public test that many other measurements don't: comparison with the election results.

The risk of abandoning the method

Some voices suggest that polling firms should abandon probabilistic in-person surveys and replace them with faster, cheaper, or more attention-grabbing methods, such as online surveys without adequate controls. The problem is that not everything that produces numbers produces knowledge.

Colombian firms that continue to use the probabilistic in-person method do so for at least two reasons. First, because it is a proven method. Second, because it is the method required by electoral regulations. Not following it would entail not only legal risks but also very high technical risks.

The recent election provided telling examples. In a pre-election publication, Semana magazine presented a tie, with approximately 12%, between the "Consultation for Life" and the "Great Consultation for Colombia." The result was very different: the "Great Consultation for Colombia" obtained many times more votes than the "Consultation for Life." It was also predicted that the "Consultation for Life" would receive more votes than the "Consultation for Solutions," and the opposite occurred.

Errors of this magnitude are not mere minor discrepancies. These are signs that the instrument used may not be suitable for measuring the phenomenon it is intended to measure. In electoral matters, speed cannot replace representativeness. The convenience of asking questions online does not replace the discipline of constructing a probabilistic sample.

Obeying the law and using a robust method is not a sign of being outdated. Nor is it an inability to conduct online surveys. It is a sign of respect for the nature of the phenomenon being measured.

Photo: Pexels | Taken from razonpublica.com

In politics, ignoring the polls is often costly.

Polls have something in common with vaccines: many people choose not to believe them, but reality often ends up contradicting their wishes.

Politics is full of leaders convinced they possess a strength that the data doesn't reflect. They believe their well-attended rallies, active social media presence, or the enthusiasm of their followers are sufficient proof of a national wave. But the ballot box is frequently unforgiving.

The case of Roy Barreras is illustrative. For many, he had the makings of a major presidential candidate. He himself stated on various occasions that he could garner millions of votes. However, the polls showed him with around 1% or less of the vote. In the end, he obtained approximately 230,000 votes, a figure much closer to what the polls had been indicating than to the expectations proclaimed by his campaign.

Roy Barreras is a leader of character, with a significant track record and an important contribution to the country's political life. But their case offers a harsh lesson: willpower doesn't replace measurement. In politics, disregarding polls might be emotionally comfortable, but it's strategically costly.

Something similar happens when campaigns that haven't yet secured a clear majority convince themselves they'll win in the first round. Politics needs enthusiasm, undoubtedly. But when enthusiasm loses its connection to reality, it ceases to be energy and becomes self-deception.

Fake news, suspicion, and erosion of trust

We live in a particularly toxic information ecosystem. Social media and artificial intelligence have made it possible for anyone to clone voices, fabricate images, spread fake news, and amplify slander. Never before has it been so easy to generate confusion on a massive scale.

In this environment, polls and polling firms also become targets of attack. It is said, for example, that they favor a particular candidate. But the methods, audits, traceability of fieldwork, and quality protocols are precisely designed to prevent this.

It is also suggested that a firm can manipulate results because it contracts with one client or another. This assertion ignores the very logic of the business. A polling firm lives off its credibility. It can survive an explained technical error; it cannot survive a loss of trust. Those who sell reliable information cannot afford to contaminate it.

The phrase attributed to Lincoln remains relevant: you can fool all of the people some of the time, and some of the people a long time, but you cannot fool all of the people all of the time. Colombian polling firms that have been in the market for thirty or forty years would not have survived without maintaining an essential commitment: to bring the voice of citizens to the public sphere and to decision-making centers without distorting it along the way.

Polls are not oracles. They do not replace politics, they do not substitute for citizen judgment, and they do not predict the future with absolute certainty. But they are among the best tools available to methodically approach what a society thinks at a given moment.

Destroying their credibility without sufficient technical justification does not strengthen democracy. It weakens it. Because a democracy needs votes, yes, but it also needs reliable ways to listen to, measure, and understand citizens before they definitively speak at the ballot box.

About the Author:

Carlos Lemoine

President of the National Consulting Center. Civil Engineer and Mathematician from the National University of Colombia. Master's and PhD in Applied Mathematics from the University of Maryland. Former Director of Studies at the Ecuadorian Data Company, former Director of Socioeconomic and Statistical Studies at the Colombian Data Company, former Director General of Socioeconomic Analysis at DANE (National Administrative Department of Statistics), and former President of the Colombian Mathematical Society.

Article taking of razonpublica.com https://razonpublica.com/encuestas-electorales-metodo-la-sospecha-la-evidencia/