The illusion of precision: How election polls shape (and mislead) voters

Polls offer a veneer of certainty, masking complex, shifting human behaviour with seductive but misleading statistics.

First published in the Mandarin.

Election polls are a staple of modern political campaigns. We are assaulted by competing polls offering a glimpse into the future fortunes of political parties and candidates.

The effects of politicians’ campaign successes and gaffes are presented in mystifying charts, and commentary focused on shifts into positive and negative ‘territory’.

To the untrained eye, the percentage point variance often seems marginal and not deserving of the accompanying feverish commentary.

In recent elections, both domestically and internationally, polls have gained a reputation for inaccuracy instead of predictive reliability. In response, pollsters and the media have adopted a hyper-analytical approach, inundating audiences with statistical jargon to bolster the perceived credibility of their findings and predictions.

Despite unreliability, limitations in the methods, and politicians' denials that they are ‘poll-driven’, election polls remain a central feature of media reporting and public interest.

Peter Dutton’s swift exit from the campaign promoting public servants’ return to the office is reportedly attributable to polls indicating that this policy lacked public support among key voter demographics, specifically women.

In 1985, Neil Postman released his influential book, ‘Amusing Ourselves to Death: Public Discourse in the Age of Show Business’. The main ideas of his work are evident in election campaigns, where it appears that public interest prioritises style over substance, as seen with tweets and viral TikToks (e.g., Anthony Albanese falling off a stage).

Today, political debates have evolved into entertainment, captivating audiences with clever quips rather than focusing on policy discussions. Election polls serve as emotional triggers aimed at swaying public opinion.

If they are increasingly unreliable, why do polls still hold such power? Beyond statistical treatments, what assumptions of human behaviour are they based on? Do they report on voter intentions or actively influence people’s decisions at the ballot box?

Fragile assumptions

The best election polls are derived from a representative sample of Australians. Marketing focus groups help explore key issues. Insights gained from these small groups can guide poll questions and provide a deeper understanding of public sentiments about an issue or candidate.

The media presents polls as reliable and scientific tools and election polling can lead the way in survey-based psychometrics and statistical methods.

However, beneath the surface, election polls rely on shaky assumptions about human behaviour, social dynamics, and voter psychology. Additionally, the assumptions that underpin the statistical techniques can also restrict the extent to which predictions can be made from the data.

At their core, polls are based on the belief that a small, well-chosen sample can reflect the views of millions. This first assumption is that the sample of the population surveyed truly represents the population.

The difference is that on election day, it’s a census of sentiments expressed as a clear decision at a voting booth.

A poll is like overhearing a lively discussion from three rooms away while wearing earmuffs. The conversation moves quickly, and the attitudes and opinions of the participants are shaped and reshaped as they contribute and listen to the views of others. Importantly, you only catch the voices of those speaking, not those who sit quietly and say nothing. When the conversation ends, you have a limited and possibly biased understanding of what people will do.

Polls rely on people’s honesty in their responses. Politics is a deeply personal topic, and many Australians are taught to keep their opinions private and avoid direct questioning—a ‘don’t ask, don’t tell’ mindset.

For a sample to be representative, pollsters must understand the views of quieter individuals; otherwise, they risk only hearing the most vocal opinions.

When do voters decide?

Polls also assume that people know how they will vote. But when do voters make up their minds?

Regardless of the issues, some people vote out of habit or loyalty. Younger voters, learning to swim through the sea of election campaign information and issues, are working out how they will make a decision.

Some people make their decisions while standing at the election booth, faced with the names of local candidates who didn’t receive media coverage. Others may change their minds after a debate, a scandal, or after overhearing the views of a trusted friend or family member.

Many may finally decide while waiting in the queue at the local school, contemplating a future that includes a sausage in bread.

This may be especially accurate now in our divided, partisan, and social media-influenced political environment.

Election campaigns are unstable. A candidate’s mistake, a viral moment on social media, or a poorly accepted policy—like Peter Dutton’s ‘work from home’ strategy—can shift a voter’s opinion in minutes or hours rather than weeks.

Why are we surprised when the polls are wrong?

When polls miss the mark, as we’ve seen recently, it's not always due to a technical or statistical mistake. Instead, they explore a complicated range of hypotheses regarding human decision-making behaviour.

The more we trust polling's predictive power, the more likely we will be blindsided when it falls short. The media frequently presents poll data with commentary that enhances its significance in the minds of readers or listeners. The numbers lend scientific weight to conjecture.

Belief in polling has been influential enough to foster a culture of ‘herding’ among polling firms, leading to adjustments or suppression of data that challenges prevailing views.

Following the 2019 federal election, Lonergan Research admitted that it refrained from publishing a poll that differed from existing ones due to concerns about potential mockery. Ultimately, the unpublished survey turned out to be the only accurate one.

Pollsters gather data—what people say they think, feel, or intend—and build statistical models based on observations from within the data and patterns of behaviour from previous elections. If specific demographics leaned one way before, it’s assumed they’ll do so again.

However, voters alter their opinions, respond emotionally, inaccurately report their intentions, or interpret identical information in various ways.

Voter attitudes evolve frequently in ways that do not reflect the data and more swiftly than pollsters can adjust their modelling.

Election polling relies on yesterday’s opinions and intentions to predict a specific future. It does so in a landscape of human behaviour—a landscape that, particularly during an election, is arguably more volatile and context-dependent than usual.

Election Day is the only accurate poll

Voters are encouraged to treat polls as forecasts rather than fragile snapshots. Elections are a collective decision shaped by millions of minds in a swirl of information, identity, and emotion. In the end, polls valiantly attempt to quantify the unquantifiable.

The act of voting is fundamentally human, messy, and unpredictable. Voters, journalists, and political strategists should treat polls as what they truly are: provisional, partial, and prone to error—not oracles of our democratic fate.