Learning how to poll is hellish.
Subscribe now for unlimited access.
$0/
(min cost $0)
or signup to continue reading
I spent a semester trying to get my head around it in my first year as a PhD student, and I'd like to give a little shout-out to Ben Goldsmith at ANU who taught me exactly how important it is to create a really good sample. A little bit of this and a little bit of that, and don't forget to weight for variables such as education, age and gender, so you get a good representation of the whole. And any errors I've just made are my own fault. Don't blame Ben, now a professor of politics.
So I have always had a lot of faith in polling for that reason. I could see how it was meant to be. I saw the science behind it. And most of the time, to my mind, it worked. Mostly. I was shocked by Brexit and I cried after the last US election.
So Wednesday morning, when it looked like the polls had predicted a Biden win but we were in fact seeing a Trumpslide, I lost my mind. How could the polling for two consecutive presidential elections be so wrong? I kept telling friends that polling was borked, that the reporting of polls was broken, and I had no idea how we were ever going to get an insight into the minds of voters.
Then I saw a grumpy tweet from Campbell White, the chair of the Australian Polling Council. I'd never heard of the APC, but I wanted to know how he could account for the catastrophe of this latest polling misjudgment. What would his mea maxima culpa sound like?
And here we are, late Thursday and it turns out that the pollsters weren't that wrong after all. Unless Trump succeeds in doing a Trump and frightening the rest of us stupid.
At first, he too was surprised by the early high vote for Trump. But instead of going to bed miserable like the rest of us, he stayed awake until 4am. That's when the results began to look a little more like the polls, so he went to sleep. In the meantime, he'd seen quite a bit of pollster-hating from the general public.
Which reassured him, in a way. In September this year, he and seven other representatives of polling organisations across Australia, such as YouGov, Ipsos and Essential, thought they'd had enough of being maligned by us regular punters and commentators. Joke. Actually they wanted to do at least three important things: one, make their processes transparent to the public; two, have standards across the industry; and three, they wanted journalists to get their reporting of polls right. Hurtful.
Try this quick quiz: What is a margin of error? What is statistical uncertainty?
Pollsters have done a shocking job of explaining why they don't get things dead right. And now they plan to fix that.
Of course, there have been some pretty big stuff-ups. The 2019 Australian federal election. The 2015 Queensland state election 2015. Yes, counters White, but that was two lots of elections in the past 20 years - everything else was within the margin of error.
So I called Peter Brent, long-time observer of polls, distant from the industry, calm and considered. Surely he would see my view! Nope, nope, nope. Basically, he told me to be patient and to recognise that most of the time the polls were correct. There have been a few high-profile exceptions, the ones we all get worked up about, but mostly the polls are right.
Sigh. I guess it is too much to expect votes to be consistent throughout the count. Shaun Ratcliff, a political scientist at the University of Sydney's United States Studies Centre, explained the strangeness of Florida to me. There was I thinking that a state full of old people, all at risk of COVID-19, and also with many people of colour, would surely vote against President Trump.
READ MORE:
Ratcliff, with Leah Ruppanner at the University of Melbourne and Andrea Carson of La Trobe, had done their own very illuminating poll. They asked 700 Americans in May whether they were concerned about COVID-19, and then asked them the same question in September (it was actually more than 700 people originally, but those were the ones that stuck around). Democrats? Yep, they understood the risks in May, and their concern remained high in September. Republicans? Not so much. In fact the higher the incidence of COVID-19 in the US, the less worried Republicans became, dropping to about 10 per cent of them being concerned. Ratcliff also reminded me that while Florida did have a high number of people of colour, there was huge variation within the Latino community - including Cubans who wouldn't vote for a Democrat in a pink fit.
The good news is that the bigger the sample, the more likely the poll is to be correct, so that's a lesson for all of us. As Ratcliff says, seat-level polling is more likely to be incorrect (the message here is not to worry about those small bits but to concentrate on the overall picture).
And there is another lesson. The next time a polling organisation rings you or emails you or texts you, you need to participate. So many of us ignore them, but the more of us respond, the more likely we are to reflect what will happen next.
Polls frighten us and sometimes they comfort us. We keep checking and looking (please follow @GhostWhoVotes on Twitter, it's fabulous for those who are addicted to polling) but they can only be right if we pitch in when the call comes.
As for the US, the last expert I spoke to said she thought Biden would get to 300 electoral votes. Which would make the polls dead right.
- Jenna Price is an academic at the University of Technology Sydney and a regular columnist.