Poll position: Will Prime Minister Bill Shorten win a second term?

Staff Writer 9 Dec 2021
6 mins
Instructions for marking your ballot

After defeating Scott Morrison in the 2019 election, Bill Shorten has had a lot on his plate with bushfires and pandemics but now, buoyed by Hillary Clinton’s second term win last November, his focus is shifting to the upcoming election.

During the campaign he can be expected to point to a relatively strong local economy in the face of the pandemic. Hopefully, for the Labor party, this will be enough to help older voters forget their rage at the scrapping of franking credits last June.

But, rather than sound economic management, Mr Shorten can thank strong pre-pandemic global trade and the strength of our trading partners, such as the EU, for our current position. Indeed, after years of economic stability following Brexit’s defeat, the EU economy is arguably the best-placed economy in the world right now.

No, you haven’t stepped into an episode of Marvel’s What If…?

The above scenario is what the world would look like if pre-election polls had been correct in the lead up to recent major elections. Bill Shorten’s Labor would have won in 2019, Clinton would have defeated Trump in 2016 and Brexit would never had happened.

These examples highlight an emerging trend of actual election results veering away from polls’ predictions, and it’s happening all around the world.

For whom the bell polls

What’s the big deal – aren’t polls just a forecasting tool anyway? A mug-punters’ form guide for elections?

While the impact of polls being off the mark may not be as severe as credit rating agencies getting it wrong in 2006, it is not as benign as Robert Walls getting it wrong 2018.

Former Prime Minister Tony Abbot knows the impact polling can have better than anyone. Malcolm Turnbull cited polling results as one of the reasons Mr Abbot should be turfed as PM in 2015.

“We have lost 30 Newspolls in a row, it is clear that the people have made up their mind about Mr Abbott’s leadership,” Mr Turnbull said at the time taking some big leaps in terms of both inference and extrapolation.

Fast forward to April 2018 and the Coalition lost 30 more polls on the trot with Mr Turnbull steering the ship. Six odd months later Mr Turnbull was himself replaced by Scott Morrison as PM. Fast forward a further five months to Mr Morrison winning a Federal election which the polls called in Mr Shorten’s favour.

It’s not just coalition leaders who have been victims. Julia Gillard was replaced by Kevin Rudd in part due to internal Labor polling which indicated he was the better choice of leader. Writing in The Guardian shortly after, Ms Gillard said she would like to see party leaders picked based on “the identification of the top new ideas – not just who is top of the opinion polls.”

So, if the polls got the 2019 election wrong, how many of the fateful 60-odd polls going against the Coalition leaders were wrong as well? Was Australia secretly happy with Mr Abbott’s onion-eating, knighthood-distributing ways? Did we wish to keep Mr Turnbull, his National Energy Guarantee and his bomber jacket in the top job? Would Australia be better off with Ms Gillard’s ‘top new ideas’ selection process?

Of course, poll results can also influence election results themselves through a ‘bandwagon effect’.

Undecided voters may just resign themselves to vote for whomever is ahead in the polls believing they will win anyway or to be on the ‘winning side’.

Poll with the punches

The first ever poll was believed to have been taken by the Harrisburg Pennsylvanian newspaper in 1824 when they asked voters their opinion on the upcoming Presidential election. Since then, they have been mostly useful and relatively accurate tools for forecasting election results. In fact, the decline in the usefulness and accuracy of polls is a relatively new phenomena, but what has caused them to lose their shine?

First, let’s address the usefulness. Over time polling has influenced democratic politics to a point where competing parties adapt policy and messaging to voters according to what they think voters want.

Poll results have become news stories in and of themselves with media outlets polling on everything from voting intentions to opinions on current issues and party leadership.

Arguably this has led to the erosion of the provision of good government and leadership. There is a risk that policy making becomes less about hard decisions on what is best for the country or what aligns with party ideologies, and more about what will edge leaders ahead in the polls.

If rival political parties are looking at similar polling results and adjusting policy accordingly you are going to have converging policy and parties will become less differentiated.

A square peg in a round poll

In terms of accuracy, there are two main drivers – the rise of mobile phones and the decline of people’s willingness to answer surveys.

Traditionally, pollsters would ring a bunch of landline phone numbers at a particular time of day when they knew most people would be home to conduct surveys. This could be expected to yield a fairly good cross section of demographics. Ring landlines these days and you will likely get older people or businesses. Most mobile phone users will not pick up an unknown number and will even register on the Do Not Call Register.

Out of those who do answer, most will hang up immediately when asked to do a survey or at best make an excuse as to why now is a bad time.

Research by The Pew Research Centre has shown that 36% of people called would agree to be polled in 1997, compared to 9% in 2016. This change could correlate with a general decline in manners or today’s busy lifestyles, but that’s a discussion for another day.

Another potential driver is what is known as ‘herding’. That is, polling agencies aligning their results to competitors’ so they aren’t making an opposite call, which could be wrong. There are obvious business reasons for doing this, if you’re a polling outlet calling it wrong and your competitors are getting it right you won’t get much more work, but it is unclear how prevalent the practice is.

There are other minor drivers which could yield wrong poll results: people change their minds after taking a poll, language barriers and some people simply lie.

Setting achievable polls

So polling is not useful or accurate and it can lead to bad government and have undue influence. So why do we still use it and what can we do to make it better?

The good news is polling agencies are aware of these issues and are trying to solve them. In 2015, Newspoll stopped surveying people by landline phones and has begun using different methodologies.


Polling agencies are now using technology to call random numbers and ask the questions through pre-recorded messages. The main advantages of robopolling are that it is much cheaper, and you can, in theory, obtain a larger sample size.

On the downside, robopolling suffers from most of the same flaws as traditional phone polling. Plus, people are more likely to hang up on a robot and in some countries automated dialling services are prohibited from calling mobile phone numbers.

Online polling

Surveys are often conducted online rather than over landlines. Obviously, this casts the net wider than using phones, but you still encounter the issue of people’s willingness to take part. Most people are time-poor, and would prefer to use their time online to shop, work, watch TikTok or all those at once.


Polling firms are also using complicated algorithms and statistical tools  that “weight” survey results and try to achieve a more accurate forecast. While these may be useful, more accurate and make up for data shortcomings, it raises questions about transparency.

Like any dataset, if polls are being manipulated the ultimate consumers of the results should be informed of the manipulation and reassured it is not being manipulated to be false or misleading.

It is also costly and time consuming to carry out this type of modelling, so it is still not used very widely.

Social media scanning

Programs exist that can gauge the number of mentions of a particular issue, leader or party on social media pages. The programs are sophisticated enough to pick up and measure sentiment as well.

The weakness of these tools is that you are only measuring people who have posted. As a general rule, these are strong supporters of a particular party who are unlikely to consider an alternative come election time. That is, they are not swing voters and they will not decide an election.

Analysing Google search trends could be a better way of gauging sentiment among swing voters and those not publicly advocating for a particular party or leader.

Poll searching

So, should we trust the polls? Maybe. They may struggle to call close elections accurately but with the large wins they’re not normally wrong. Then again, most people can see a landslide coming themselves.

Do we need the polls? Maybe. Polls are reassuring and imply political predictability and stability. Most humans prefer certainty to chaos, especially the business community, retirees, graduates, investors and politicians themselves.

We can help

Thankfully the Purple team does not need to trust the polls or back a winner in an election. As a bipartisan firm, we focus on providing clients with the best lobbying service and government relations advice, regardless of which party is in government.


About the author

Cameron is a Government Relations Advisor with a unique and varied background, in both the public sector and private sector. He is adept at assisting private companies to navigate government processes, policy and approvals. His vast experience is invaluable in helping translate the languages of government and industry.

More Purple news: