Pollsters should bin SAS and go down the pub

[shaunbhoy]

Old Greg is NOT a disguised employee

But he IS Brave Sir Robin!!

[Old Greg]

Well - they navigated it less badly than everyone else.

They still said that Trump was unlikely to win, but this was better than everyone else which said that Trump didn't have a cat-in-hell's chance.

He was unlikely to win, so fivethirtyeight was correct.

[centurian]

Yes, fivethirtyeight navigated this one well.

Well - they navigated it less badly than everyone else.

They still said that Trump was unlikely to win, but this was better than everyone else which said that Trump didn't have a cat-in-hell's chance.

[ZARDOZ]

Who needs polls when you have a tool that has correctly predicted every result since 1980?
The Keys to the White House: Forecast for 2016 | National Council for the Social Studies

[jamesbrown]

Yes, fivethirtyeight navigated this one well.

Yep. It was interesting watching those conditional probabilities change as the results came in. Clinton was still favourite until quite late on, but I think his model reflected the undecided vote better than others, as evidenced by the large spread throughout the night until things finally tightened up. It was revealing that, immediately prior to the vote, the possibility of a marginal Clinton win, a huge Clinton win, and a marginal trump win were roughly equally likely

[Old Greg]

It really depends on the polling. The results was well within the margin of error for the popular vote. It wasn't good in swing states, but that's tougher to poll. Some of the national polls and models did pretty well on the electoral college too (USC/LA Times on the polling and Nate Silver gave a reasonably high chance of a Trump upset on the polling plus model). Remember, there were a lot of declared undecided voters, way more than usual, so this wasn't an easy one to predict.

Yes, fivethirtyeight navigated this one well.

[jamesbrown]

It really depends on the polling. The results was well within the margin of error for the popular vote. It wasn't good in swing states, but that's tougher to poll. Some of the national polls and models did pretty well on the electoral college too (USC/LA Times on the polling and Nate Silver gave a reasonably high chance of a Trump upset on the polling plus model). Remember, there were a lot of declared undecided voters, way more than usual, so this wasn't an easy one to predict.

[woohoo]

Could it also be down to:

- People just saying what they feel is the right thing to say
- Pollster not polling wider sample
- Pollsters need to work out a way to include comments/trends seen on social websites

Well half of me agrees with you but the other 60% thinks it just doesn't add up.

[jbond007]

The pollsters had a gap in the last polls which where in the margin of error.

Could it also be down to:

- People just saying what they feel is the right thing to say
- Pollster not polling wider sample
- Pollsters need to work out a way to include comments/trends seen on social websites

[SueEllen]

The pollsters had a gap in the last polls which where in the margin of error.

[Fronttoback]

How could the pollsters have misread the mood of the man on the street so badly? Maybe they are spending too much time in front of a computer doing clever stuff and not enough time talking to voters in the pub- where alcohol allows voters to speak their minds freely- thus reflecting the outcome better.