The Solution for Polling Accuracy: Less Logic, More Xenophobia!
By Nick Drew
So, it’s happened again. After the British general election in 2015 and the Brexit referendum, now comes the latest blow to the reputation of the polling industry with Trump’s ‘unexpected’ win in the US presidential race. And, as ever, the opprobrium has already started, with the world seemingly placing the blame at the feet of those pesky pollsters. “Ohhhh, they got it wrong again!”, “Can’t they do anything right?!”.
It’s enough to make me wish I’d chosen a different career; one where I could quietly do a slapdash job, safe in the knowledge that when my and my colleagues’ failings came to light, nobody would care, nor suggest that they know better than us. Something like working in a telephone help centre; being a quality control checker on German diesel cars; or a rocket scientist. Around SpaceX’s latest rocket pre-takeoff explosion, there seemed to be quite a lack of people asking really, what are they all doing, it’s not that hard.
But after this latest crushing blow to the research industry’s credibility, and assuming that people are right – that observing the polling figures differently would have changed the result of the election – what’s the problem?
Well, there’s a clear trend of unconscious observer bias. A recent WSJ article demonstrated how the same set of polling figures can lead to quite different conclusions, with the specific predicted outcome depending upon the statistical models and personal interpretations applied by a pollster. And unconscious bias plays a large role in this.
Researchers are fairly smart people: educated, with a head for numbers, reasonably articulate, and able to understand the idea of a multi-cultural world. They’re also generally employed, and the industry is becoming, on average, younger and more female over time. But these very attributes are inherently limiting, and shape the way researchers think. Polling firms didn’t predict a vote for Brexit because to any logical person, the idea of the UK seceding from its continent is utterly ridiculous. Likewise, a Trump victory wasn’t widely predicted because the idea of a xenophobic bigot winning the most important job in the world through a popular vote is unfathomable – at least to logical, educated, reasonably articulate people who can understand the idea of a multi-cultural world.
So what’s to be done? Fortunately, the answer is clear and, indeed, easy. In order to break out of this limited mindset, and become better at predicting elections and referenda, research firms need to have greater diversity in their ranks. Forget women and ethnic minorities (those are so last year): the research industry needs to be employing more angry, old white people; those who didn’t finish school; men who like to grab women by the unmentionables; those who don’t like to talk through their problems, but instead want to rage at how the system is fixed. Most importantly, perhaps, we need to do better at hiring people who think that foreigners are to blame for everything, and for whom a weird mix of national isolationism and imperialism provides the ideal solution to all the world’s problems. Only then can polling firms break away from the tyranny of the logical approach, and start to better understand and more accurately predict the views of the electorate.
Nick Drew is VP, Strategy & Insights at Fresh Intelligence. The views above are his own and not intended to be taken seriously.