After 2016, it’s understandable why media outlets are taking care to emphasise that even polls showing Joe Biden clearly ahead of Donald Trump in the national vote are not reasons to think that Biden is definitely going to win. It’s looking very likely, but very likely is not the same as certain.
But it is certainly possibly to overdose on the difference between very likely and certain. Enter stage left, The Independent, with a new poll showing Biden ten points ahead of Trump, the sort of lead that even the sort of polling error we’ve seen in the past and unlucky breaks in electoral college would still seen Biden elected.
The good bit of the story is the quote from the pollster at the end of the story:
James Johnson, a former Downing Street pollster and JL Partners founder, says time could be the former vice president’s biggest ally with just two weeks of campaigning left.
“With two weeks to go, Donald Trump is running out of time. After all the trials and tribulations of the last few weeks, Biden continues to hold a clear 10-point lead with likely voters,” he said, “a margin which would take him into the White House even in the event of significant polling error.”
But now let’s consider the headline:
2020 polls: Biden leads Trump by 10 points – but polling finds warning signs for Democrat
Exclusive: Former VP is lagging behind Hillary Clinton’s support among white college-educated and Black voters with two weeks to go, Independent poll reveals
This is a classic example of how news stories about polls – and other data, such as unemployment statistics or inflation figures – first present the overall picture, then given an outlier, and then discard the truth and talk only about the outlier as if the outlier is the overall picture.
In this case: the overall picture is Biden is 10 points ahead. Then the outlier – there are two groups of voters where he is doing worse than Clinton. But – the crucial unstated bit, omitted from not only the headline but the rest of the story – Biden is further ahead overall than Clinton was in 2016. So if he’s under-performing compared with her in two groups of voters, he’s also over-performing among far more voters. That doesn’t get mentioned in the story.
What’s worse, the story is then written as if only the under-performing is happening, with all the greater over-performing ignored, with comments such as, “the poll found ample reasons for Mr Biden and his campaign team to worry”. No mention of how he’d doing better than Clinton overall with voters. Even the one brief reference to Biden doing better with one group of voters is immediately dismissed:
The survey shows Mr Biden doing better with Latino voters than Ms Clinton did four years ago. He has a 57 per cent lead, compared to her 38 per cent. But in another possible warning for team Biden, JL partners adds this caveat: “Note that the September poll showed Trump performing slightly better than in 2016.” There are indicators Mr Biden is underperforming compared to a Democratic nominee who lost the Electoral College race to Mr Trump,
You really wouldn’t guess from that the truth of Biden doing better than Clinton overall, would you?
It’s similar to how stories about, say, inflation will often mention things that have risen by more than the average but ignore those which have risen by less than the average. The final result from such stories is that it’s very easy to be misled by them, because all that discussion about things that are out on one side of the average makes it sounds like reality isn’t where the average is. More on that in non-polling stories in my book, Bad News: what the headlines don’t tell us.