Polling is back in vogue during the Voice referendum debate and those results — rather than the substance of the debate — could influence the way some vote. Australia’s view on the Indigenous Voice referendum has been keenly centred on political polling. News outlets regularly fixate on poll results, which they treat as crystal balls forecasting the referendum outcome, months in advance. Political polling in the debate reflects a self-fulfilling prophecy, where the numbers become the story rather than being used to help tell it. Polling is centre stage in Australia now, but only four years ago it was on the brink: the country’s second-largest media company, Nine Entertainment, declared it would stop publishing weekly polls from Ipsos after the 2019 federal election result defied pollsters’ predictions. But polling has surged back to prominence amid a narrative of declining support for the Yes campaign, which extends with each fortnightly poll update. Different outlets approach the same poll data with their own twists: The Guardian reports each new Essential poll as a “slim majority” for the No vote, while Sky News Australia coverage of Roy Morgan results says support for a Yes vote “plunges“, “freefalls“, and “plummets“. This renewed attention on political polls is not surprising, given their potency within the political and media ecosystem. Polls — especially those conducted regularly — provoke reaction and are big business, with complex, often obscured long-term relationships between news outlets and pollsters. Backlash to polls in Australia came after the 2019 federal election, in which federal Opposition leader Bill Shorten lost an “unlosable election” after spending two years ahead in polling. The disparity between polling and the outcome led some commentators to describe the failure to predict Shorten’s loss as “worse than the Trump polling fail” in 2016. Criticism centred on the polls’ accuracy, but also on the strong expectation for a result that did not eventuate that these polls helped to create. In 2019, statisticians and scientists, notably including Nobel Laureate physicist Brian Schmidt, called for major reviews of pollsters’ methodologies, warning of “undue, or undeserved, influence” on vote results. A group of pollsters — YouGov, Essential, and uComms — responded by forming the Australian Polling Council to reassure journalists and the public of the quality of their data and improve accuracy in reporting. But worries about poll accuracy in reporting distract from more critical issues — like how polls are used to further existing political narratives, and the impact of reporting poll results in the first place. Schmidt’s commentary focussed on the likelihood of pollsters adjusting— probably unintentionally — poll results to match each other. But this concern should extend beyond just those conducting the polls. The media too readily overlooked the effects of constantly polling and publishing poll results to the public. The most effective use case for polling data is often internally, providing political campaigns with regular profiles of the voters they are fighting to win. This doesn’t mean they should be published more widely: if the aim of the media is to inform the public, the question must be asked what benefit there is from knowing how others will vote. Aspects of the Voice referendum might generate increased attention to political polls — and generate greater susceptibility to simple narratives. The increased focus on polling could be due to the heightened stakes in any referendum, given its binary nature. A referendum takes a weighty issue and funnels it into two opposite responses. But, unlike an election, referenda require a majority of states to approve the constitutional change, as well as a national majority. Reporting that highlights the fact that the national figures do not reflect the full picture, and that there are still a substantial number of uncertain voters, might be missed entirely. The ABC and the Guardian maintain multi-poll trackers to try to produce a more powerful picture of campaign performance beyond the limits of individual polls. But in doing that, they remove undecided respondents, “to give a forced yes/no response”. This is the poll reporting trade-off: a clearer image, but one constructed by removing nuance. This can have an unchecked effect on how we perceive the campaign. Beyond technical issues with getting polling right, it’s hard to regulate against people using polls out of context. Some Australian voters say they don’t understand the constitutional changes or feel the involvement of non-Indigenous Australians in this issue is fundamentally unethical. In any case, there could be more demand from people seeking cues on how to vote.