Following the first Clinton-Trump Presidential debate this week, Trump supporters were taking to twitter, citing online polls as evidence that their candidate performed more strongly, often under the hashtag #TrumpWon. Trump came out on top in a number online polls – so he won, right?
Not quite. To understand online polls, you have to understand how they work. Most online polls are used as a form of promotion, often with a prize as incentive. As one example, in 2012 school textbook company Chegg ran an online promotional competition, with the prize being $10,000 for the top five schools, and a Taylor Swift concert for the winner. The obvious intent here is for schools to encourage their students and their families to enter the contest in order to get more resources for the school, with the use of a major performer used as a flashy and attention-grabbing gimmick to enthuse students.
4chan, a community of online pranksters, had other ideas. An idea gained traction, that it would be funny to use their online influence to send Taylor Swift to perform at a school for the deaf. Boston’s Horace Mann School for the Deaf and Hard of Hearing was chosen, and 4chan members flooded the vote to back this school in such huge numbers that it gathered mainstream media coverage.
This had a secondary effect. A volunteer who’d worked with deaf children brought up a time they’d helped organise a concert for deaf children. As a result, a secondary campaign was set up to get Taylor Swift to Horace Mann, in order to raise awareness of the fact that deaf children can enjoy concert music with some slight modifications.
The motivations were a little complex, but essentially there were far more people who became aware of the contest with the intent of voting for Horace Mann than those who entered with the intent of voting for their own local school. As a result, it was pretty much inevitable that Horace Mann would win the overall poll. (They did, and Swift sent an extra cash prize to Horace Mann rather than perform there.)
This process has parallels in online political polls. Fringe parties will always be given a smaller amount of mainstream media coverage than mainstream parties, meaning that, while voters in general will feel they have enough information on the mainstream parties, potential supporters of fringe parties will need to go searching through online sources and social media for the information they want. As a result, fringe parties will generally have a disproportionately high online presence, often with strong communities dedicated to supporting the party. As members of the Green Party of England and Wales we have firsthand experience of this, and saw, during the run-up to the 2015 election, a lot of these online communities.
When the online edition of a newspaper article ends with a poll on the bottom of the page, members would often share a link, with the explicit instruction of telling people to click through specifically to take part in the poll. There is some tactical logic for this – the poll could be used to ‘show’ how strong support for the party is, and convince undecided voters that the lesser party is more of a viable option than it actually is.
But this can lead to supporters kidding themselves about how popular the party they support is amongst the wider population. We ran into a number of UKIP supporters who seemed to sincerely believe that their party would win an overall majority in the House of Commons, citing online polls as ‘proof’. (In the end UKIP won 1 of 650 seats.)
Following the first Clinton – Trump debate, a ‘Polls Compilation Megathread!’ was posted on a Donald Trump supporting Reddit messageboard. Reddit’s ‘The_Donald’ board has over 200,000 subscribers, and as it’s heading informs visitors that “posts supporting other candidates will be removed” it’s fair to say that there’s a pro-Trump bias to the board.
There was also a large number of links to polls on 4chan, with one user going as far as saying that
The actual merits of the two candidates’ debating actually means nothing. In the eyes of millions of normalfags, the “victor” will be the one with the best post-debate polls.
It’s up to us to hand Trump the victory.
Another user pointed out that
YOU CAN VOTE MULTIPLE TIMES ON THE TIME.COM POLL USING INCOGNITO MODE
So the intention to rig the online polls wasn’t even subtle. And yet, at the end of the night, Trump supporters spread their blatantly rigged polls, with 11 high-profile sites in particular being cherrypicked to support this idea.
So just as 4chan hijacked a poll intended to promote a contest’s sponsors in return for investment in schools, using it instead to mock Taylor Swift; this week Donald Trump supporters used polls intended to measure the level of support for election candidates, to give the impression that their favoured candidate performed more strongly than he did.
The counter-argument to this is a high-profile poll conducted by CNN. Prior to the debate, CNN phoned participants randomly to ask them to take part afterwards. The pre-debate polling reveals that participants in the poll, overall, had a Democrat bias going in. Trump supporters with a large social media presence have been using to support the idea that CNN’s poll was rigged.
Let’s think about the details. Pre-debate figures show the Democrat Party were represented by 41% to the Republican Party’s 26%, so going in Clinton had a roughly 1.5 to 1 advantage over Trump. After the debate, this lead had increased to a 62% to 27% lead – roughly a 2.3 to 1 advantage in favour of Clinton.
CNN hasn’t released a full breakdown of movement among those they polled, so we can’t be certain what effect the performances had – how many moved from Independent to Democrat, from Independent to Republican, from Republican to Democrat, or Democrat to Republican. We can see however, that, in this group of people:
- Clinton’s lead amongst this group of people increased from a 50% greater support to 130% greater support than Trump has
- While 22% moved away from Independent, Trump picked up 1% net, Clinton picked up 21% net
CNN will have been trying to strike a balance, between gathering as large a sample size as possible, and keeping them as balanced as possible. We know how unbalanced voters in the poll were going in, so we can bear that in mind when judging how much we trust or distrust the CNN poll. It’s the rate of change more than the headline result which is the most informative – and that figure is dramatic in Clinton’s favour.
Yes, the CNN poll only surveyed 521 people, so it’s a much smaller sample size – some of the online polls received hundreds of thousands of clicks. And yes, there was a bias going in. But we know how biased they were, and so we can take conclusions away from the poll.
On the online polls, however, we have no way of knowing the biases the larger number of clickers had going in. (The sites in question may be able to say how many people clicked onto the relevant page from Reddit, but even that wouldn’t be definitive proof either way.)
So, in short, the online polls for which a greater number of people clicked on Donald Trump’s name merely show that a greater number of people clicked on Donald Trump’s name. Beyond that, they’re meaningless. Despite this, other Trump supporters with large social media presence have been repeating the propaganda that the CNN poll was ‘rigged’.
Similarly to the CNN poll, a focus group of undecideds ran by Republican pollster Frank Luntz ended with 16 of 22 participants coming down on Clinton’s side. The majority of other professionally ran polls have concluded that the debate was a strong win for Clinton.
The democratisation of power and information that the internet provides is, on the whole, a great thing. But it’s important to be aware of the methodology behind attention-grabbing announcements, rather than repeating news that tallies with what we already believe.
Online polls in which Trump supporters click on Trump’s name multiple times don’t show that he performed well in the debate, any more than the Chegg poll shows that the students of Horace Mann are massively passionate Taylor Swift fans.