They had her anywhere between a 70-90% chance to win. If you predict 90% chance that something will happen, and it always happens, your prediction is wrong because you should have predicted 100%.
When I hear someone say “you can’t trust the polls because they got 2016 ‘wrong’” they are just telling me they don’t understand statistics.
And its important to note that these predictions were for the pop vote, which she did actually win, so they were actually right.
I’m not sure this is entirely true. Many polls just look at the popular vote, but most of the polls that claim “chance of winning” take into account the EC.
538 had her going into the election with a 70% chance of winning the electoral college. Nate Silver also went on multiple shows basically doing everything he could to get people to understand that meant 3 out of 10 times she loses.
No, 538 (and RCP?) actually has a rolling projection of ‘real’ chance to win the EC. But the chances of Hillary declined from >90% to 70% in the last week or so. When she was >90% everybody would say it looked like she was going to win, and that’s what people remember.
It’s been awhile since I read anything about that, but it seems like the last time I read about it, was something along: “80% of polls have Hillary projected to win”, but the actual polls that they were using were all almost within the margin of error.
People in almost never speak about the margin of error when presenting a poll, especially one that’s favorable to them.
f you look at the fine print, and see the margin of error percentage, then you apply the maximum amount to both people in the race, you’ll see a lot of times it’s a tie.
I understand the point you’re making about probabilities, but we’re speaking in the context of politics. Polls accurately predicted the results in 2008 and 2012. Something fundamentally changed in 2016, and the polls were off across the board.
And in reality we are talking about an weighted error difference of about 1.3/1.5 points between 2008/2012 and 2016. It’s not like they got it massively wrong.
You said something changed, and then I showed how last election they were actually more accurate than in the past. And I already pointed out that they were “wrong” about 2016. So I’m not sure what your point is about this.
they are just telling me they don’t understand statistics.
You’re right, but in fairness to the regular person who gets their news from regular news outlets, they were being told that Clinton had a 98% chance of winning when in reality it was more like 75%. The fact is while everyone was cocky in 2016 and nervous in 2020 I was the opposite because I followed the polls and Biden in 2020 had consistently bigger leads on Trump than Clinton in 2016 with even bigger leads in swing states. His odds of winning were much greater than hers and the likely margin of victory was much higher, but they were being underestimated by a media machine that was absolutely snakebit after going all in on congratulating HRC in June for being the first woman president with a dem supermajority in both houses of congress and flipping Texas blue.
Funny that this was in response to me and not the above poster that claimed that something happened in 2016 that made them no longer reliable.
Additionally, I suspect you don’t really know what you are talking about because the issue you point out is not a statistical issue, but that they are just not a good measurement to begin with. Which isn’t even a good point either because they do a pretty good job of consistently getting pretty close. In the last election the mean error was only about 4.3 and they didn’t seem to favor either side.
Polls would be ok if the sample was peefectly random. However it is never fully random, and in practice they always overrepresent politically active people and underrepresent the poor.
They had her anywhere between a 70-90% chance to win. If you predict 90% chance that something will happen, and it always happens, your prediction is wrong because you should have predicted 100%.
When I hear someone say “you can’t trust the polls because they got 2016 ‘wrong’” they are just telling me they don’t understand statistics.
And its important to note that these predictions were for the pop vote, which she did actually win, so they were actually right.
I’m not sure this is entirely true. Many polls just look at the popular vote, but most of the polls that claim “chance of winning” take into account the EC.
538 had her going into the election with a 70% chance of winning the electoral college. Nate Silver also went on multiple shows basically doing everything he could to get people to understand that meant 3 out of 10 times she loses.
No, 538 (and RCP?) actually has a rolling projection of ‘real’ chance to win the EC. But the chances of Hillary declined from >90% to 70% in the last week or so. When she was >90% everybody would say it looked like she was going to win, and that’s what people remember.
Oh yeah, the Comey Probe. Back in the days when having the FBI open an investigation into you was enough to kill your presidential aspirations.
Or at least that was the case for Hillary Clinton and the moderate voter bloc, but somehow Donald Trump is not held to such high standards.
It’s been awhile since I read anything about that, but it seems like the last time I read about it, was something along: “80% of polls have Hillary projected to win”, but the actual polls that they were using were all almost within the margin of error.
tl;dr 80% had Hillary winning by about 2-3%.
People in almost never speak about the margin of error when presenting a poll, especially one that’s favorable to them.
f you look at the fine print, and see the margin of error percentage, then you apply the maximum amount to both people in the race, you’ll see a lot of times it’s a tie.
I understand the point you’re making about probabilities, but we’re speaking in the context of politics. Polls accurately predicted the results in 2008 and 2012. Something fundamentally changed in 2016, and the polls were off across the board.
And “polls were historically accurate” in 2022.
And in reality we are talking about an weighted error difference of about 1.3/1.5 points between 2008/2012 and 2016. It’s not like they got it massively wrong.
Pollsters gave Hillary a 90% chance to win, likely because of over-representation of college graduates in polling, as well as a late shift in support in swing states.
You said something changed, and then I showed how last election they were actually more accurate than in the past. And I already pointed out that they were “wrong” about 2016. So I’m not sure what your point is about this.
You’re right, but in fairness to the regular person who gets their news from regular news outlets, they were being told that Clinton had a 98% chance of winning when in reality it was more like 75%. The fact is while everyone was cocky in 2016 and nervous in 2020 I was the opposite because I followed the polls and Biden in 2020 had consistently bigger leads on Trump than Clinton in 2016 with even bigger leads in swing states. His odds of winning were much greater than hers and the likely margin of victory was much higher, but they were being underestimated by a media machine that was absolutely snakebit after going all in on congratulating HRC in June for being the first woman president with a dem supermajority in both houses of congress and flipping Texas blue.
I know right? Some people haven’t played XCOM, and it shows.
What are you talking about? Polls are not valid statistics, they are riddled with biases that can’t be eliminated.
Funny that this was in response to me and not the above poster that claimed that something happened in 2016 that made them no longer reliable.
Additionally, I suspect you don’t really know what you are talking about because the issue you point out is not a statistical issue, but that they are just not a good measurement to begin with. Which isn’t even a good point either because they do a pretty good job of consistently getting pretty close. In the last election the mean error was only about 4.3 and they didn’t seem to favor either side.
Polls would be ok if the sample was peefectly random. However it is never fully random, and in practice they always overrepresent politically active people and underrepresent the poor.