A Twitter post lists the results of five polls which asked people who won last week’s ITV leaders debate between Jeremy Corbyn and Boris Johnson. The post quickly went viral.
Four of the polls found Jeremy Corbyn to be the winner, yet the tweet claims that the BBC and ITV only reported the YouGov poll—in which 51% of peoples said Johnson was the winner (after excluding the 7% of people who answered “don’t know”).
This may sound like a scandal, but it isn’t. As BBC journalist Joey D’Urso says: the tweet “sounds scientific but it's utterly wrong”.
The reason is that the four polls which found Mr Corbyn to be the winner are not representative of the general population. A good survey, by design, asks a group of people (the “sample”) what they think, and tries to make sure that those people represent the wider population.
All four polls which found Mr Corbyn to be the winner were conducted on Twitter. That means anybody with a Twitter account could answer it, and there’s no way of knowing if those who did are representative of the views of the wider public.
And they probably weren’t representative—only 17% of the UK aged 16 or over uses Twitter, and those people are typically younger, wealthier, and more urban than the general population.
By contrast, the YouGov poll was conducted with a sample that represented the general public.
The BBC reported the YouGov poll as suggesting “the public were evenly split on who had won the debate”, while ITV said “A YouGov snap poll suggested 51% of Britons believed Mr Johnson won the debate compared to 49% for Mr Corbyn. Those who answered "don't know" were removed from the result, with YouGov adding the figures are so close as to be within the margin of error.”
It’s better to include those who answered “don’t know”, if you want an understanding of perceptions among the public as a whole. Doing this, 48% thought Mr Johnson was the winner, 46% thought it was Mr Corbyn, and 7% answered don’t know.
Twitter and Facebook polls are poor indicators of public opinion, as we’ve written before.