Politics

How Did The Pollsters Get The Federal Election So Wrong?

Let's talk about polls, how they work, and why the media gets them wrong.

election preferences, senate voting, election

Chances are, they probably didn’t. I know it seems wild — everyone said Labor would win the federal election on the weekend, then they didn’t.

Here’s the thing though, polls aren’t a prediction or a projection. They’re literally a record of what a group of people think at a moment in time. Their usefulness relies entirely on asking the right group of people the right question at the right time — and in modern life this is actually super hard.

A Brief History Of Polling

A bit of background on polls. Polling first came along — like most weird election innovations — in the USA. A group of people in a town in Pennsylvania indicated who their preferred candidate for President was in 1824 — and this ‘straw poll’ reflected the national result. A novelty! It was a stunt for the local newspaper, who sold lots of copies, and then this cute newspaper-selling novelty proliferated around the country. Thus, polling was born.

As you’ve likely guessed though, this was hardly scientific. It’s not scientific because the group of people saying who they were going to vote for weren’t representative of the whole voting population. They also weren’t random, they just opted in. In polling, this group of people is called a sample, and a sample should always reflect the spread of demographics of the population whose opinion you are trying to measure. In the case of elections, that’s the whole voting populace. In other cases, it might be members of a particular organisation or people considering buying a car — it all depends on what you want to know and who you want to know about.

National polls take days to run, usually over the phone or online, to get a random and representative sample. People who complete a poll usually receive an automatic phone call or an email invitation, and they choose to complete the poll or not. But it’s pretty hard in the days of mobile phones (consider the demographics of people who still have landlines — hardly representative!), increased multilingualism, and a growing population to make sure that your sample is both random AND representative.

In order to have a poll to publish, the last poll before an election is usually taken about at least a few days before election day itself. That’s up to a whole week of people still deciding their vote. When a poll is taken, anyone who says that they don’t know, refuse to answer, or are unsure are removed from the total. It makes sense, you can’t just force these people to make a decision or to tell you — and you can’t just randomly allocate them (you’re just guessing then). But almost all of these people will make up their mind at some point, potentially affecting the outcome massively if they all go largely the same way.

Now imagine that these people who we’re leaving out because they’re not telling us anything, all actually have something in common. Maybe in this election, the people who made up their minds very late — and so were excluded from the polls — were all people who don’t care about politics. They don’t care, they weren’t paying attention, they made up their mind on the way into the booth on Saturday, a full three or four or five days after the polling stopped.

So What Are Polls Actually Good For?

This isn’t about making excuses for why the polls didn’t ‘predict’ the outcome (they aren’t designed to, remember). If polls aren’t predictive, what are they good for? Making headlines.

Polls are a good tactic for getting your issue, your candidate, or your campaign a good story in the media. Let’s say, for example, you want more action on climate change. You commission a poll that indicates most voters agree with your proposition. You give the results to a media outlet or two, who publish them with a bit of added political context, and voila! You get some headlines that help to push your agenda forward.

No reputable polling company will let a dodgy poll be published in their name, or design a poll to give a particular outcome, that’s genuinely not in their interests. But if a poll is positive, then it’s in the interests of the campaign to let people know that – because humans like to do what everyone else is doing. If you look like you’re going to win, people are more inclined to support you and that improves your chances of winning! Around we go again.

But polls have to be seen in context. They’re not infallible, and we should stop expecting them to be. Some people genuinely are making up their mind right up until election day. Some segments of the population are more difficult to find to even ask their opinion in the first place. Maybe there’s even a group who are just so disengaged they can’t be bothered answering — and maybe they’re more collectively inclined to vote a certain way (it’s hard to be sure though, they don’t really like doing surveys).

Elections are a contest. They should first and foremost be a contest of ideas, not likeability or popularity. An over-focus on polls drives us away from looking at the big problems and analysing possible solutions towards what seems popular — and that’s even worse for democracy than a few weird polls.


Hayley Conway is a writer and campaigner with over a decade of experience in strategy and movement building. She currently works as a consultant for Essential Media, and was previously a senior adviser to an Greens leader Richard Di Natale and spent four years in the USA working on national and international campaigns.