I've been talking to friends in the US and they all say the same thing, Biden leads the polls but the feeling on the ground is that Trump has the momentum because he's been gifted a stronger hand by the riots and the Dems lack of condemnation of their own side and the very blatant media bias. This is among all likely Dem voters as well.
The CNN graphic of "Fiery but mostly peaceful protests" in front of burning buildings has become a recruiting tool for Trump among middle class suburban whites, it's team Trump saying "we told you, they'll lie to your face and expect you to accept it" and it's working for them among parents of my friends who were previously in the Dem column.
That's the narrative, and it's one that - for example - Conrad Black and Andrew Sullivan have both written about in the last two weeks.
But I'm a numbers rather than a narrative kind of guy. Let's see if voters are actually changing.
One problem with the polls last time was that in order to make the samples balanced, they compromised on randomness, which meant they failed to pick up on important movements.
A lot is being made of people overcompensating for the 2016 shock by overstating Trump's chances but I'm beginning to wonder whether a closer analogy is 2017 with Biden as May with regards to the polls.
As @MaxPB has pointed out, plus others such as Andrew Sullivan and Conrad Black, people do not like the violence and the fear it is coming to a town near you and it is starting to become a major issue. @rcs1000 pointed out Republican registrations in Florida are running ahead of the Democrats but my understanding is the Republicans have eroded the Democrat's lead since 2016 in PA as well, which also bodes ill for Biden.
More to the point, both the campaign's actions and what is being seen on the ground doesn't seem to match the polls. Trump has just opened an office in Minnesota and he has visited New Hampshire.
Trump is repeatedly banging the drum linking Biden with the violence and Biden is now running adverts condemning the violence. This is where the NY Times states about the adverts:
"The Biden campaign said the ad would air nationally on cable television and in local markets in nine battleground states: Arizona, Florida, Michigan, Minnesota, Nevada, North Carolina, Ohio, Pennsylvania and Wisconsin."
That doesn't sound like a campaign that is confident it is high single digits ahead.
Ultimately, though, Biden's national polling numbers aren't changing.
Pundits and stories and narrative have very little predictive power compared to hard numbers.
The problem I have is whether the data we are getting from these polls is clean.
If I am dealing with a set of financial numbers on which to make a forecast, I can assume they are clean in that they have been audited, subject to oversight etc.
That is not the case here. As many on here have pointed out, US polling methodology is not as audited as in the UK. More concerning is that the polls are seen as both sides to push their positions so they become used as a political football and used as to rally the troops so it is entirely possible you are getting self-selected samples.
Also, as mentioned on a previous thread, both parties will have continual internal polling going on which is why I am a great believer in looking at their actions as that demonstrates their "true" view.
I think the idea the campaigns have any idea what's going on is not backed by the evidence. Hillary had no idea she was losing in the rust belt. .
Hilary's campaign did know they were deep in trouble.
I've been talking to friends in the US and they all say the same thing, Biden leads the polls but the feeling on the ground is that Trump has the momentum because he's been gifted a stronger hand by the riots and the Dems lack of condemnation of their own side and the very blatant media bias. This is among all likely Dem voters as well.
The CNN graphic of "Fiery but mostly peaceful protests" in front of burning buildings has become a recruiting tool for Trump among middle class suburban whites, it's team Trump saying "we told you, they'll lie to your face and expect you to accept it" and it's working for them among parents of my friends who were previously in the Dem column.
That's the narrative, and it's one that - for example - Conrad Black and Andrew Sullivan have both written about in the last two weeks.
But I'm a numbers rather than a narrative kind of guy. Let's see if voters are actually changing.
One problem with the polls last time was that in order to make the samples balanced, they compromised on randomness, which meant they failed to pick up on important movements.
A lot is being made of people overcompensating for the 2016 shock by overstating Trump's chances but I'm beginning to wonder whether a closer analogy is 2017 with Biden as May with regards to the polls.
As @MaxPB has pointed out, plus others such as Andrew Sullivan and Conrad Black, people do not like the violence and the fear it is coming to a town near you and it is starting to become a major issue. @rcs1000 pointed out Republican registrations in Florida are running ahead of the Democrats but my understanding is the Republicans have eroded the Democrat's lead since 2016 in PA as well, which also bodes ill for Biden.
More to the point, both the campaign's actions and what is being seen on the ground doesn't seem to match the polls. Trump has just opened an office in Minnesota and he has visited New Hampshire.
Trump is repeatedly banging the drum linking Biden with the violence and Biden is now running adverts condemning the violence. This is where the NY Times states about the adverts:
"The Biden campaign said the ad would air nationally on cable television and in local markets in nine battleground states: Arizona, Florida, Michigan, Minnesota, Nevada, North Carolina, Ohio, Pennsylvania and Wisconsin."
That doesn't sound like a campaign that is confident it is high single digits ahead.
Ultimately, though, Biden's national polling numbers aren't changing.
Pundits and stories and narrative have very little predictive power compared to hard numbers.
The problem I have is whether the data we are getting from these polls is clean.
If I am dealing with a set of financial numbers on which to make a forecast, I can assume they are clean in that they have been audited, subject to oversight etc.
That is not the case here. As many on here have pointed out, US polling methodology is not as audited as in the UK. More concerning is that the polls are seen as both sides to push their positions so they become used as a political football and used as to rally the troops so it is entirely possible you are getting self-selected samples.
Also, as mentioned on a previous thread, both parties will have continual internal polling going on which is why I am a great believer in looking at their actions as that demonstrates their "true" view.
I think the idea the campaigns have any idea what's going on is not backed by the evidence. Hillary had no idea she was losing in the rust belt. Trump and his advisors thought they had lost. Nigel Farage conceded defeat immediately after EU Referendum polls closed.
Your polling methodology point is a good one. It is entirely possible - maybe even likely - that there are shy Trump supporters, who are not being picked up.
The issue for the President is that (on current polls) there needs to be an unprecedented level of polling error.
And there's one last issue. Why isn't Biden's share moving? If the riots were having an effect, he should still see his share move. Because changes in opinions should be picked up, even if the overall totals are wrong. And we're not seeing any meaningful move in the Biden total.
Trafalgar is rated as a C- pollster by FiveThirtyEight with an average polling error of 5.8 points. Best ignored
Yes ignore the only pollster who predicted Trump would win Michigan and Pennsylvania in 2016 if you wish but don't come crying to me on election night if you are wrong
What’s your latest prediction for the election?
Still fence-sitting so you are certain you can claim you were right?
It is not fence sitting when I say it will be between 260-275 votes either way.
If Biden wins with a landslide obviously I will have been miles out and if Trump won comfortably that would also be the case even if less likely
Looking on from afar, that doesn't add up. A tie is 269, seats over that on one side have to be matched by exactly the same number under if they apply either way.
So no more than 275 (+6) over implies no fewer than 263 (-6) under, whereas you say 260.
Alternatively, no fewer than 260 under implies no more than 278 over, whereas you say only 275.
OK, 263 to 275 for either candidate then if you really want to pin me down.
That is my prediction
At this point, if I limited myself to a spread of 12 to match yours, I would predict Biden getting between 279 and 291.
I would go with Biden 338-350 for a current prediction in a 12 EV band. If Texas flips then 400+ is on the cards.
I've been talking to friends in the US and they all say the same thing, Biden leads the polls but the feeling on the ground is that Trump has the momentum because he's been gifted a stronger hand by the riots and the Dems lack of condemnation of their own side and the very blatant media bias. This is among all likely Dem voters as well.
The CNN graphic of "Fiery but mostly peaceful protests" in front of burning buildings has become a recruiting tool for Trump among middle class suburban whites, it's team Trump saying "we told you, they'll lie to your face and expect you to accept it" and it's working for them among parents of my friends who were previously in the Dem column.
That's the narrative, and it's one that - for example - Conrad Black and Andrew Sullivan have both written about in the last two weeks.
But I'm a numbers rather than a narrative kind of guy. Let's see if voters are actually changing.
One problem with the polls last time was that in order to make the samples balanced, they compromised on randomness, which meant they failed to pick up on important movements.
A lot is being made of people overcompensating for the 2016 shock by overstating Trump's chances but I'm beginning to wonder whether a closer analogy is 2017 with Biden as May with regards to the polls.
As @MaxPB has pointed out, plus others such as Andrew Sullivan and Conrad Black, people do not like the violence and the fear it is coming to a town near you and it is starting to become a major issue. @rcs1000 pointed out Republican registrations in Florida are running ahead of the Democrats but my understanding is the Republicans have eroded the Democrat's lead since 2016 in PA as well, which also bodes ill for Biden.
More to the point, both the campaign's actions and what is being seen on the ground doesn't seem to match the polls. Trump has just opened an office in Minnesota and he has visited New Hampshire.
Trump is repeatedly banging the drum linking Biden with the violence and Biden is now running adverts condemning the violence. This is where the NY Times states about the adverts:
"The Biden campaign said the ad would air nationally on cable television and in local markets in nine battleground states: Arizona, Florida, Michigan, Minnesota, Nevada, North Carolina, Ohio, Pennsylvania and Wisconsin."
That doesn't sound like a campaign that is confident it is high single digits ahead.
Ultimately, though, Biden's national polling numbers aren't changing.
Pundits and stories and narrative have very little predictive power compared to hard numbers.
The problem I have is whether the data we are getting from these polls is clean.
If I am dealing with a set of financial numbers on which to make a forecast, I can assume they are clean in that they have been audited, subject to oversight etc.
That is not the case here. As many on here have pointed out, US polling methodology is not as audited as in the UK. More concerning is that the polls are seen as both sides to push their positions so they become used as a political football and used as to rally the troops so it is entirely possible you are getting self-selected samples.
Also, as mentioned on a previous thread, both parties will have continual internal polling going on which is why I am a great believer in looking at their actions as that demonstrates their "true" view.
I think the idea the campaigns have any idea what's going on is not backed by the evidence. Hillary had no idea she was losing in the rust belt. Trump and his advisors thought they had lost. Nigel Farage conceded defeat immediately after EU Referendum polls closed.
Your polling methodology point is a good one. It is entirely possible - maybe even likely - that there are shy Trump supporters, who are not being picked up.
The issue for the President is that (on current polls) there needs to be an unprecedented level of polling error.
And there's one last issue. Why isn't Biden's share moving? If the riots were having an effect, he should still see his share move. Because changes in opinions should be picked up, even if the overall totals are wrong. And we're not seeing any meaningful move in the Biden total.
If they posted their crosstabs they would be a lot easier to analyse.
Nevertheless, their simple "error rate" - i.e. the average amount they were wrong for Dem and Republican candidates - in 2018 was not great, at 5.6 percent.
“When we talk about hidden voters, what we’re talking about is the social-desirability bias, and that is when people basically tell a live-caller what they think will get them judged least harshly,” says Cahaly. “Some races have that, some races don’t.” Cahaly points to the Florida 2018 gubernatorial race as an example. Democrats had branded Republican Ron DeSantis as a racist, and all pollsters, except Trafalgar, showed Democrat Andrew Gillum with a lead in the days before the election. DeSantis won the election by 0.4 points.
“In 2020, the presidential question has social-desirability bias,” says Cahaly. He tries to capture the views of “hidden voters” in several ways. “I start with a voter file that has everything from occupation to incomes to education levels to voting history to likely religion,” says Cahaly. “We give people multiple ways to participate in our polls,” he adds. “We do live calls, we do automated calls, we do texts, we do emails, we do other digital platforms.” Cahaly tries to get a sample of at least 1,000 respondents in any statewide poll: “Big samples are better samples.”
He thinks a lot of pollsters for media outlets simply ask too many questions — if a pollster is asking 30 questions, “you end up with people who really care too much about politics,” says Cahaly. Trafalgar’s surveys ask “no more than nine [questions] and [take] less than three minutes to complete.”
Yes, I've linked to that article myself, about why we shouldn't ignore Trafalgar.
But let me turn it around. Trafalgar didn't do very well in 2018, where they applied these same weightings.
Now, it could be that 2020 is just like 2016. Or it could be that it's more like 2018.
The difference between 2016 and 2018 though was Trump was not on the ballot, many Trump voters in 2016 stayed home in 2018. The question is if they will turn out for him again in 2020 and whether other pollsters are capturing them or missing them unlike Trafalgar
Didn't the GOP congressional vote outperform Trump nationally in 2016?
Campaign to get Britain back to work flounders. Divisions remain over office social distancing, home working in Whitehall and lack of a slogan. Boris Johnson’s drive to get workers back to the office has been postponed amid warnings that the Government’s own social distancing guidelines prevent firms from getting all of their staff back to their desks..
I've been talking to friends in the US and they all say the same thing, Biden leads the polls but the feeling on the ground is that Trump has the momentum because he's been gifted a stronger hand by the riots and the Dems lack of condemnation of their own side and the very blatant media bias. This is among all likely Dem voters as well.
The CNN graphic of "Fiery but mostly peaceful protests" in front of burning buildings has become a recruiting tool for Trump among middle class suburban whites, it's team Trump saying "we told you, they'll lie to your face and expect you to accept it" and it's working for them among parents of my friends who were previously in the Dem column.
That's the narrative, and it's one that - for example - Conrad Black and Andrew Sullivan have both written about in the last two weeks.
But I'm a numbers rather than a narrative kind of guy. Let's see if voters are actually changing.
One problem with the polls last time was that in order to make the samples balanced, they compromised on randomness, which meant they failed to pick up on important movements.
A lot is being made of people overcompensating for the 2016 shock by overstating Trump's chances but I'm beginning to wonder whether a closer analogy is 2017 with Biden as May with regards to the polls.
As @MaxPB has pointed out, plus others such as Andrew Sullivan and Conrad Black, people do not like the violence and the fear it is coming to a town near you and it is starting to become a major issue. @rcs1000 pointed out Republican registrations in Florida are running ahead of the Democrats but my understanding is the Republicans have eroded the Democrat's lead since 2016 in PA as well, which also bodes ill for Biden.
More to the point, both the campaign's actions and what is being seen on the ground doesn't seem to match the polls. Trump has just opened an office in Minnesota and he has visited New Hampshire.
Trump is repeatedly banging the drum linking Biden with the violence and Biden is now running adverts condemning the violence. This is where the NY Times states about the adverts:
"The Biden campaign said the ad would air nationally on cable television and in local markets in nine battleground states: Arizona, Florida, Michigan, Minnesota, Nevada, North Carolina, Ohio, Pennsylvania and Wisconsin."
That doesn't sound like a campaign that is confident it is high single digits ahead.
Ultimately, though, Biden's national polling numbers aren't changing.
Pundits and stories and narrative have very little predictive power compared to hard numbers.
The problem I have is whether the data we are getting from these polls is clean.
If I am dealing with a set of financial numbers on which to make a forecast, I can assume they are clean in that they have been audited, subject to oversight etc.
That is not the case here. As many on here have pointed out, US polling methodology is not as audited as in the UK. More concerning is that the polls are seen as both sides to push their positions so they become used as a political football and used as to rally the troops so it is entirely possible you are getting self-selected samples.
Also, as mentioned on a previous thread, both parties will have continual internal polling going on which is why I am a great believer in looking at their actions as that demonstrates their "true" view.
I think the idea the campaigns have any idea what's going on is not backed by the evidence. Hillary had no idea she was losing in the rust belt. Trump and his advisors thought they had lost. Nigel Farage conceded defeat immediately after EU Referendum polls closed.
Your polling methodology point is a good one. It is entirely possible - maybe even likely - that there are shy Trump supporters, who are not being picked up.
The issue for the President is that (on current polls) there needs to be an unprecedented level of polling error.
And there's one last issue. Why isn't Biden's share moving? If the riots were having an effect, he should still see his share move. Because changes in opinions should be picked up, even if the overall totals are wrong. And we're not seeing any meaningful move in the Biden total.
If they posted their crosstabs they would be a lot easier to analyse.
Nevertheless, their simple "error rate" - i.e. the average amount they were wrong for Dem and Republican candidates - in 2018 was not great, at 5.6 percent.
“When we talk about hidden voters, what we’re talking about is the social-desirability bias, and that is when people basically tell a live-caller what they think will get them judged least harshly,” says Cahaly. “Some races have that, some races don’t.” Cahaly points to the Florida 2018 gubernatorial race as an example. Democrats had branded Republican Ron DeSantis as a racist, and all pollsters, except Trafalgar, showed Democrat Andrew Gillum with a lead in the days before the election. DeSantis won the election by 0.4 points.
“In 2020, the presidential question has social-desirability bias,” says Cahaly. He tries to capture the views of “hidden voters” in several ways. “I start with a voter file that has everything from occupation to incomes to education levels to voting history to likely religion,” says Cahaly. “We give people multiple ways to participate in our polls,” he adds. “We do live calls, we do automated calls, we do texts, we do emails, we do other digital platforms.” Cahaly tries to get a sample of at least 1,000 respondents in any statewide poll: “Big samples are better samples.”
He thinks a lot of pollsters for media outlets simply ask too many questions — if a pollster is asking 30 questions, “you end up with people who really care too much about politics,” says Cahaly. Trafalgar’s surveys ask “no more than nine [questions] and [take] less than three minutes to complete.”
Yes, I've linked to that article myself, about why we shouldn't ignore Trafalgar.
But let me turn it around. Trafalgar didn't do very well in 2018, where they applied these same weightings.
Now, it could be that 2020 is just like 2016. Or it could be that it's more like 2018.
The difference between 2016 and 2018 though was Trump was not on the ballot, many Trump voters in 2016 stayed home in 2018. The question is if they will turn out for him again in 2020 and whether other pollsters are capturing them or missing them unlike Trafalgar
Didn't the GOP congressional vote outperform Trump nationally in 2016?
Trump got a higher voteshare in 2016 than the Republicans did at the 2018 congressional elections
I've been talking to friends in the US and they all say the same thing, Biden leads the polls but the feeling on the ground is that Trump has the momentum because he's been gifted a stronger hand by the riots and the Dems lack of condemnation of their own side and the very blatant media bias. This is among all likely Dem voters as well.
The CNN graphic of "Fiery but mostly peaceful protests" in front of burning buildings has become a recruiting tool for Trump among middle class suburban whites, it's team Trump saying "we told you, they'll lie to your face and expect you to accept it" and it's working for them among parents of my friends who were previously in the Dem column.
That's the narrative, and it's one that - for example - Conrad Black and Andrew Sullivan have both written about in the last two weeks.
But I'm a numbers rather than a narrative kind of guy. Let's see if voters are actually changing.
One problem with the polls last time was that in order to make the samples balanced, they compromised on randomness, which meant they failed to pick up on important movements.
A lot is being made of people overcompensating for the 2016 shock by overstating Trump's chances but I'm beginning to wonder whether a closer analogy is 2017 with Biden as May with regards to the polls.
As @MaxPB has pointed out, plus others such as Andrew Sullivan and Conrad Black, people do not like the violence and the fear it is coming to a town near you and it is starting to become a major issue. @rcs1000 pointed out Republican registrations in Florida are running ahead of the Democrats but my understanding is the Republicans have eroded the Democrat's lead since 2016 in PA as well, which also bodes ill for Biden.
More to the point, both the campaign's actions and what is being seen on the ground doesn't seem to match the polls. Trump has just opened an office in Minnesota and he has visited New Hampshire.
Trump is repeatedly banging the drum linking Biden with the violence and Biden is now running adverts condemning the violence. This is where the NY Times states about the adverts:
"The Biden campaign said the ad would air nationally on cable television and in local markets in nine battleground states: Arizona, Florida, Michigan, Minnesota, Nevada, North Carolina, Ohio, Pennsylvania and Wisconsin."
That doesn't sound like a campaign that is confident it is high single digits ahead.
Ultimately, though, Biden's national polling numbers aren't changing.
Pundits and stories and narrative have very little predictive power compared to hard numbers.
The problem I have is whether the data we are getting from these polls is clean.
If I am dealing with a set of financial numbers on which to make a forecast, I can assume they are clean in that they have been audited, subject to oversight etc.
That is not the case here. As many on here have pointed out, US polling methodology is not as audited as in the UK. More concerning is that the polls are seen as both sides to push their positions so they become used as a political football and used as to rally the troops so it is entirely possible you are getting self-selected samples.
Also, as mentioned on a previous thread, both parties will have continual internal polling going on which is why I am a great believer in looking at their actions as that demonstrates their "true" view.
I think the idea the campaigns have any idea what's going on is not backed by the evidence. Hillary had no idea she was losing in the rust belt. .
Hilary's campaign did know they were deep in trouble.
IIRC part of the campaign knew they were in trouble. The group around Hillary knew better than all those silly old fools with their polling....
Comments