experts warned that a Trump win was still possible.

But only if polls were even worse than they were in 2016.

But look at where Biden is right now. He's up by 10 points.

More than that - look at that. He's above 50 percent.

Hillary Clinton was well below 50 percent.

So did history repeat itself in 2020?

Let's compare the polling averages listed by FiveThirtyEight with the actual results.

Both nationally and in every battleground state, Trump outperformed the polls again.

That ended up being decisive in only two states: Florida and North Carolina,

where the winner was different than what was projected,

but didn't change the overall electoral outcome.

But the stark difference between polls and reality in states like Iowa, Ohio and Wisconsin

still stands out.

One pollster who predicted Trump's wide margin in Iowa correctly,

against conventional wisdom,

was Ann Selzer.

It takes a little bit of practice to live in what I call 'the corral of pot shots'

while your poll is out there waiting to see what happens with the election.

Polling Trump in midwestern states such as Iowa

has been a riddle for the industry ever since 2016.

So what makes her methods different?

There are a lot of pollsters who do what I call they poll backwards,

that is they look at what happened in previous elections

and use that to adjust their sample of today to hopefully project an electorate of tomorrow.

And I don't do that. I take nothing from the past and build it into my polls.

Therefore, I hope, the future can reveal itself to me in my data.

Part of the problem these days, according to Rachel Bitecofer,

is who is available to pollsters.

There is this perpetual problem of finding people from the bottom 50 percent,

left and right, who are going to do a survey.

They don't know who Mitch McConnell is.

They don't know who Nancy Pelosi is.

They don't read The New York Times.

They don't watch the news.

What makes things even harder is that Republicans in the US

are much more likely to distrust news organisations who sponsor a lot of polling.

Which could be one explanation of why Trump underperforms in polls.

We're in this time period where especially the political right

has found value in getting people to distrust government,

and getting people to divest themselves from civic participation.

So what could be a solution?

The only way that pollsters will ever be able to measure them

is by asking longer questions and by empowering these people, telling them

that if you speak to us the world is going to hear you.

It actually requires a different questionnaire than what we normally would do with voters.