Don’t Believe the Polls – Australian Edition

Remember how the polls said there was no chance Hilary could lose the 2016 election? Well they’ve done it again. This time in Australia. The expert who predicted Trump, Brexit – and Scott Morrison

The referenced article, or any of the others on the Australian elections, isn’t really of interest. What is of interest is the polling was wrong. Completely wrong. Again.

Despite trailing in every major poll for three years, the Coalition retained power on Saturday night and could yet form majority government.

With polling and betting markets missing the mark, experts are increasingly turning to social media to judge voter sentiment on a larger scale.

And with the crackdown on conservative thought by Big Social Media, even that will be hard to pin down.

So the next time some talking head, or politician, says that X percent of the population favors their latest Big Government power grab, ask them about 2016, or Australia.

So I’ve been trying to figure out how polling companies can get so much wrong. So I went looking for their methodology. Pew Research was my first (and only) stop. What I found is inexplicable. Trump’s Staunch GOP Supporters Have Roots in the Tea Party: Methodology.

This isn’t the only piece of Pew’s methodology I looked at. And what I saw wasn’t 100 percent uniform, but a whole bunch of what I saw, fell into this pattern.

The ATP [Pew Research Center’s American Trends Panel] is a nationally representative panel of randomly selected U.S. adults recruited from landline and cellphone random-digit-dial surveys.

They will determine that 312-555-12XX is in a particular part of the Chicago area (for example) and dial random digits. They do similar calling for blocks of numbers that are cellphones. I can’t think of a more self-selecting group to answer questions. If you “randomly” call my number, and you aren’t in my contacts list, I’m going to assume you are the credit fraud scammer, and ignore the call. (Anyone with real business will leave a message.) If Pew research left a message – which I don’t believe they do, but I’m not 100 percent clear on that – I wouldn’t call them back. Would you?

Maybe dialing people randomly could have worked back in the dark ages, before answering machines, but with caller ID and voicemail, and the proliferation of scams, who answers a call from an unknown number?

And this isn’t a poll from the dark ages. This is from a document published in 2019, based on group recruited to be on this panel in 2014 and 2015.

Polls in which the surveyed population is a self-selecting sample are not usually too good. People who will answer random calls from unknown numbers. People who will agree to work with Pew for years out. I don’t know anyone, of any political affiliation, that I think would fall into a group like this. Random sample? Not hardly.

And then I’m still a fan of Mike Royko’s call to lie to exit pollsters, from 1984.

But he said if enough voters lie, ‘The entire nation will be treated to one of the finest evenings of television viewing since the tube was unleashed.

‘As the evening wears on and the actual votes are counted, we will see Dan become more and more wild-eyed. We’ll see Peter hyperventilating. And even Davie will look like he is fully awake. And they’ll all be stammering about how ‘goodness, something seems to have gone wrong.’

Surprisingly good description of what happened in 2016, even if the players had changed.