By Stuart Bruce, 29/11/2014
We need to talk about voting intention polling. It gets treated as if it means more than it does – as if it is firmly pre-empting how everyone will vote. Polling masquerades as the prediction of behaviour, something that should be left to behavioural researchers because it is notoriously hard to do. It should be regarded as a measure of current perception on any particular question only, and leave it at that. The only polls that can reliably predict future behaviour are the polls done the day before an election. 24 hours ahead of the result. Hardly worth the enthusiasm.
The main reason voting intention polling is flawed is obvious. A decision with a real effect is different to a decision with a hypothetical effect. This gap between the actual and the hypothetical cannot be closed without adding other aspects of political psychology to inform the overall picture. These aspects could make it high-level research and would make it much more reliable as a predictive measure. Yet voting intention polling is all over the place and often gets used as if it is predictive. Voting intention polling is analogous to asking someone how they would spend their money if they won the lottery. Of course, they would give most of it to charity and help their ailing third cousin pay off his mortgage. How they actually spend their money is different of course.
And voting intention polling isn’t benign either. It can have dramatic and irreversible effects. What might be framed initially as a guide becomes over-used as people base more of their conclusions on them. They manifest, albeit tacitly, as crystal balls for the election coverage. The blitzkrieg of polling occurring in Victoria at the moment only produces a quagmire of asserted causes, actions, effects and opinions. Much like a Barry Jones ‘knowledge nation’ diagram, the concepts and suppositions flow everywhere without ever pinning down causality and in no way do they neatly explain the current political landscape prior to the election.
Without even asking the questions to ascertain underlying political ideologies, moral values and other such details, it is presumptuous to assume that polling can predict behaviour. Even the name is misleading. Judging something to be an ‘intention’ mistakenly fosters the assumption that it is, even in some way, a reliable prediction of future behaviour. The person intends to do something, therefore they will. We should call it voting indication, a seemingly minor change but the implied predictive assumption is lessened. An ‘indication’ represents the temporal aspect, where a decision is made for that moment only, whereas an ‘intention’ directly asserts an outline of future behaviour.
Where should we go from here? Firstly, we should always qualify how inaccurate polls can be so people regard them as appropriately limited. And secondly, measurements (such as a political values breakdown) should be added alongside regular polling to better inform people about what voting decisions are based on. These steps would make a better representative democracy as politicians would attune their policies to people’s actual political values and not the various snapshots of transient opinion which would be exposed when they weren’t aligned with values.
Politicians would likely become less risk-averse, knowing they can argue their policies by drawing on the underlying values and beliefs of the electorate and defend themselves against the inevitable poll dip when they propose a hard but necessary reform. They could therefore avoid the career death spiral that occurs when a reform causes sliding polls which can make the polls slide even further. All opinion polling should be analysed against a backdrop of the people’s values systems, media use, engagement styles and more. Voting intention polling, which has little if any of these aspects, is just too flawed to use as a building block for any grand predictions about a looming election outcome.
Stuart Bruce is the executive director of the Centre for Applied Political Psychology