Mr. Eek, it remains a monumentally stupid decision.
"Shall we simply stick with this legendary track that's produced countless fantastic races?" "We could. Or we could cut down a forest to make an identikit shitbag circuit." "Good idea!"
It's a left hand track, right next to a crime ridden shit hole favela, the surface is fucked and they can't build enough run off in section 2. It's cheaper to build a new Tilkedrome than to fix its many issues. They will also never get MotoGP/WSBK to Interlagos either so a new one makes sense. Also tracks in American timezones are all wrong for European/Asian TV scheduling so they have to work hard to hang on to a race.
Meanwhile covid rips through the universities in the biggest controlled experiment in the history of ethically-iffy evidence based medicine - which may be a good thing, mebbe noes, but we'll know in a couple of weeks to what extent it is self-contained, and non-lethal/non-serious, in those particular petri dishes. Not sure why this has escaped comment in response to those complaining that we have given no thought to alternative strategies, risk segmentation and so on.
I suppose it might explain why the A-Level results were tweaked with such alacrity. Some of the "kidz" demonstrating back in August may now be studying the law of unintended consequences.
Compare and contrast 538 2016 - 2020. Not quite the same vibe.
To me these are the key charts. Clinton's support always looked soft, with wildly oscillating poll leads. Biden's lead looks like it is built on far firmer foundations. Americans were willing to give Trump a chance in 2016, frequently against their better judgement. In 2020 I think minds are firmly made up.
Consider the time period missing from that chart. It strongly suggests only Covid is beating Trump. The announcement of a vaccine would be expected to change everything.
Not necessarily. It looks to me like it shows Trump benefiting from a brief rally round the leader effect when the Covid crisis started hitting mid march to beginning April - as did most heads of government. Prior to that it shows Biden with a better chance - but that was when Sanders was still running.
Plus any decent model will have more uncertainty the further away you are from the election - so if things stay the same Biden's lead will equal a greater and greater likelihood of him winning as we get closer to the election.
Plus I'm not sure which chart you are using - it doesn't seem to be the 538 once quoted above.
If you look at approval ratings you probably get a more consistent picture over time, which doesn't depend on the Dem's nomination process, decreasing uncertainty as we get closer to the election, or temporary effect of the Covid rally round the leader. https://projects.fivethirtyeight.com/trump-approval-ratings/ Shows Trump doing about the same now as before the end-March Covid rally. If anything his approval ratings are slightly better now than before Covid.
The chart is from the Economist prediction model, because 538 doesn't show that far back.
It isn't a rally round the leader effect in March, because Covid didn't hit the US until later than Europe and the UK. It goes south for Trump after the deaths start piling up.
The point is that the situation before Covid was very different. Trump had a very good chance of winning. It's only Covid that has changed that. This is not what I would describe as a "firm foundation" for Biden - because a lot could be forgiven in the relief that a vaccine would bring.
This is the problem, isn't it? The app needs to be sensitive (in the sense of triggering for the vast majority of people who are in fact infected due to close contact) to be effective in limiting spread, but that likely implies - unless the algorithms discriminate spectacularly well - a low specificity (lots of triggers for people who have not in fact been infected). As people start to notice and report this on Twitter and the like, support for the app and use of it will likely fall, even if it is doing a good job at reducing infections overall.
To maintain support, there needs to be some analysis - and publicity - of how much the app is helping to reduce infections.
Actually, that's quite alarming. I understood, having raised the point and been re-assured, that if the app showed positive, then one would be contacted (fairly) quickly to establish roughly at what time the positive contact took place, especially if like Mr Waugh, one has been out and about. If, in effect, the warning is being triggered by being (let us say) back to back with someone in a pub when in fact the chance of interchange of virus is low, then as, Mr S suggests, the app is rapidly going to be come unpopular and discontinued. Or, as well as 'you need to self-isolate' it should authorise a test as soon as as.
On the one hand: Scott finding pertinent, relevant tweets and posting them here where many people don't have the time to find them directly.
On the other: paragraphs and paragraphs of people saying how they could never support Boris but they support Boris and how they want him gone but want him to stay until he has done the thing they don't support him for.
I'll take the former.
This may be a complete non sequitur, but I value both Scott and Big_G.
Having Big_G and HYFUD on the site - a former and current prominent local Tory party activist - provides a fascinating real time picture of the evolution of the Conservative Party. It feels like a case study of the application of Gresham's Law to politics.
One reason that this site is so good is the variety of political viewpoints represented, and by and large the civil nature of debate. That really is something quite unusual on the Internet and social media.
I quite like some of @Scott_P tweets, but his much rarer original author posts are very good indeed.
A bit of cooling off time is not such a bad thing though, hope @malcolmg has had his 3 match ban over soon.
As if by magic I have been pardoned, I will be on my best behaviour.
BTW and O/T @malcolmg sends his regards. He is apparently banned at the moment, not sure why.
Personally I miss his colourful contributions.
No! Malcy may be obnoxious but he`s OUR obnoxious.
Bring him back.
Thank you Stocky and hey presto I am back.
Welcome. How's Mrs M progressing?
OKC, thanks going well, CT scans showed lungs improved , and her heart rhythm has seemingly sorted itself out ( was worth delaying it seems ) so that procedure cancelled and she can come of blood thinners and beta blockers. Steroids reduced as well so apart from picking up colds etc the signs are all going well. She had tests this week for lung function and whilst they will be scarred hopefully it will show decent function. So all positive at present.
BTW and O/T @malcolmg sends his regards. He is apparently banned at the moment, not sure why.
Personally I miss his colourful contributions.
No! Malcy may be obnoxious but he`s OUR obnoxious.
Bring him back.
Thank you Stocky and hey presto I am back.
Hello again Malcolm. Didn't realize you'd been banned. I assumed you were on a break. Jealous now. Wish I could be banned. It spells kudos and street cred. Like getting the cane at school. Come out of the Head's study, cheeks red - face too - but not crying. Cheers ring out across the playground. You grin and salute. Bit of a hero. I speak from observation rather than experience sadly. I was never that boy. It was always someone else. Plus ca change.
BTW and O/T @malcolmg sends his regards. He is apparently banned at the moment, not sure why.
Personally I miss his colourful contributions.
No! Malcy may be obnoxious but he`s OUR obnoxious.
Bring him back.
Thank you Stocky and hey presto I am back.
Welcome. How's Mrs M progressing?
OKC, thanks going well, CT scans showed lungs improved , and her heart rhythm has seemingly sorted itself out ( was worth delaying it seems ) so that procedure cancelled and she can come of blood thinners and beta blockers. Steroids reduced as well so apart from picking up colds etc the signs are all going well. She had tests this week for lung function and whilst they will be scarred hopefully it will show decent function. So all positive at present.
Excellent. Very good news. You must both be greatly relieved!
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
I believe that the actual, full poll is 51% Biden.
The 21% lead stuff comes from cherry picking the data from after the Trump Covid diagnosis - it was a five day poll.
I haven't seen more detail yet, but that sounds close to people taking un-adjusted sub-samples from raw polling data, to me.
You're probably right in that.
The Boston Herald says: "In two days of polling before Trump got COVID, the president trailed Biden by just a 46%-41% margin. In the three days of polling after the coronavirus diagnosis, Biden held a 55%-34% lead. That means Biden’s lead grew by a whopping 16 points from pre-COVID to post-COVID. Among all the 1,003 registered, likely voters in the nationwide Franklin Pierce-Herald poll, Biden now holds a 51%-37% lead over Trump less than a month before Election Day."
However, the splitting of polling data shouldn't be dismissed out of hand. In the distant past. I have seen Anthony Wells run threads with heavy caveats that breaking down YouGov polling when it spans a major event, so it's not entirely spurious when analysed carefully (unlike the way the Boston Herald reported it mind). The sub-samples won't be properly weighted so there will be an extra element of randomness in there. The 16% difference here is so marked that you shouldn't disregard the strong likelihood that there was some change, even if it may have been small. The fact that the samples are reasonably balanced (40% and 60%) before and after Trump fessed up to having Covid also helps.
So I think it's reasonable to conclude that a poll showing Biden with a 14% lead contains compelling evidence that the lead would most likely have been a bit larger if the poll had been conducted to the same methodology with the same full sample to a slightly later timeframe.
I believe that the actual, full poll is 51% Biden.
The 21% lead stuff comes from cherry picking the data from after the Trump Covid diagnosis - it was a five day poll.
I haven't seen more detail yet, but that sounds close to people taking un-adjusted sub-samples from raw polling data, to me.
You're probably right in that.
The Boston Herald says: "In two days of polling before Trump got COVID, the president trailed Biden by just a 46%-41% margin. In the three days of polling after the coronavirus diagnosis, Biden held a 55%-34% lead. That means Biden’s lead grew by a whopping 16 points from pre-COVID to post-COVID. Among all the 1,003 registered, likely voters in the nationwide Franklin Pierce-Herald poll, Biden now holds a 51%-37% lead over Trump less than a month before Election Day."
However, the splitting of polling data shouldn't be dismissed out of hand. In the distant past. I have seen Anthony Wells run threads with heavy caveats that breaking down YouGov polling when it spans a major event, so it's not entirely spurious when analysed carefully (unlike the way the Boston Herald reported it mind). The sub-samples won't be properly weighted so there will be an extra element of randomness in there. The 16% difference here is so marked that you shouldn't disregard the strong likelihood that there was some change, even if it may have been small. The fact that the samples are reasonably balanced (40% and 60%) before and after Trump fessed up to having Covid also helps.
So I think it's reasonable to conclude that a poll showing Biden with a 14% lead contains compelling evidence that the lead would most likely have been a bit larger if the poll had been conducted to the same methodology with the same full sample to a slightly later timeframe.
Saying there are indication that the lead might have been bigger if the polling was all after the event is one thing.
Putting up a headline say 21% is going too far, I think.
As 2016 proved the national popular vote is irrelevant anyway, it is key swing states like PA that matter and Biden now down to a 4% lead there with Emerson today
Mr. Eek, it remains a monumentally stupid decision.
"Shall we simply stick with this legendary track that's produced countless fantastic races?" "We could. Or we could cut down a forest to make an identikit shitbag circuit." "Good idea!"
It's a left hand track, right next to a crime ridden shit hole favela, the surface is fucked and they can't build enough run off in section 2. It's cheaper to build a new Tilkedrome than to fix its many issues. They will also never get MotoGP/WSBK to Interlagos either so a new one makes sense. Also tracks in American timezones are all wrong for European/Asian TV scheduling so they have to work hard to hang on to a race.
It is fun in the sim though.
Maybe F1 should follow the example of cricket and go for a quick-fire, slimmed-down, crowd-pleasing variant for the benefit for spectators too "tired" to concentrate? Circular track, 10 laps, 80mph max.
On the one hand: Scott finding pertinent, relevant tweets and posting them here where many people don't have the time to find them directly.
On the other: paragraphs and paragraphs of people saying how they could never support Boris but they support Boris and how they want him gone but want him to stay until he has done the thing they don't support him for.
I'll take the former.
This may be a complete non sequitur, but I value both Scott and Big_G.
Having Big_G and HYFUD on the site - a former and current prominent local Tory party activist - provides a fascinating real time picture of the evolution of the Conservative Party. It feels like a case study of the application of Gresham's Law to politics.
One reason that this site is so good is the variety of political viewpoints represented, and by and large the civil nature of debate. That really is something quite unusual on the Internet and social media.
I quite like some of @Scott_P tweets, but his much rarer original author posts are very good indeed.
A bit of cooling off time is not such a bad thing though, hope @malcolmg has had his 3 match ban over soon.
As if by magic I have been pardoned, I will be on my best behaviour.
BTW and O/T @malcolmg sends his regards. He is apparently banned at the moment, not sure why.
Personally I miss his colourful contributions.
No! Malcy may be obnoxious but he`s OUR obnoxious.
Bring him back.
Thank you Stocky and hey presto I am back.
Welcome. How's Mrs M progressing?
OKC, thanks going well, CT scans showed lungs improved , and her heart rhythm has seemingly sorted itself out ( was worth delaying it seems ) so that procedure cancelled and she can come of blood thinners and beta blockers. Steroids reduced as well so apart from picking up colds etc the signs are all going well. She had tests this week for lung function and whilst they will be scarred hopefully it will show decent function. So all positive at present.
Excellent. Very good news. You must both be greatly relieved!
I believe that the actual, full poll is 51% Biden.
The 21% lead stuff comes from cherry picking the data from after the Trump Covid diagnosis - it was a five day poll.
I haven't seen more detail yet, but that sounds close to people taking un-adjusted sub-samples from raw polling data, to me.
You're probably right in that.
The Boston Herald says: "In two days of polling before Trump got COVID, the president trailed Biden by just a 46%-41% margin. In the three days of polling after the coronavirus diagnosis, Biden held a 55%-34% lead. That means Biden’s lead grew by a whopping 16 points from pre-COVID to post-COVID. Among all the 1,003 registered, likely voters in the nationwide Franklin Pierce-Herald poll, Biden now holds a 51%-37% lead over Trump less than a month before Election Day."
However, the splitting of polling data shouldn't be dismissed out of hand. In the distant past. I have seen Anthony Wells run threads with heavy caveats that breaking down YouGov polling when it spans a major event, so it's not entirely spurious when analysed carefully (unlike the way the Boston Herald reported it mind). The sub-samples won't be properly weighted so there will be an extra element of randomness in there. The 16% difference here is so marked that you shouldn't disregard the strong likelihood that there was some change, even if it may have been small. The fact that the samples are reasonably balanced (40% and 60%) before and after Trump fessed up to having Covid also helps.
So I think it's reasonable to conclude that a poll showing Biden with a 14% lead contains compelling evidence that the lead would most likely have been a bit larger if the poll had been conducted to the same methodology with the same full sample to a slightly later timeframe.
I believe that the actual, full poll is 51% Biden.
The 21% lead stuff comes from cherry picking the data from after the Trump Covid diagnosis - it was a five day poll.
I haven't seen more detail yet, but that sounds close to people taking un-adjusted sub-samples from raw polling data, to me.
You're probably right in that.
The Boston Herald says: "In two days of polling before Trump got COVID, the president trailed Biden by just a 46%-41% margin. In the three days of polling after the coronavirus diagnosis, Biden held a 55%-34% lead. That means Biden’s lead grew by a whopping 16 points from pre-COVID to post-COVID. Among all the 1,003 registered, likely voters in the nationwide Franklin Pierce-Herald poll, Biden now holds a 51%-37% lead over Trump less than a month before Election Day."
However, the splitting of polling data shouldn't be dismissed out of hand. In the distant past. I have seen Anthony Wells run threads with heavy caveats that breaking down YouGov polling when it spans a major event, so it's not entirely spurious when analysed carefully (unlike the way the Boston Herald reported it mind). The sub-samples won't be properly weighted so there will be an extra element of randomness in there. The 16% difference here is so marked that you shouldn't disregard the strong likelihood that there was some change, even if it may have been small. The fact that the samples are reasonably balanced (40% and 60%) before and after Trump fessed up to having Covid also helps.
So I think it's reasonable to conclude that a poll showing Biden with a 14% lead contains compelling evidence that the lead would most likely have been a bit larger if the poll had been conducted to the same methodology with the same full sample to a slightly later timeframe.
Saying there are indication that the lead might have been bigger if the polling was all after the event is one thing.
Putting up a headline say 21% is going too far, I think.
As 2016 proved the national popular vote is irrelevant anyway, it is key swing states like PA that matter and Biden now down to a 4% lead there with Emerson today
Well the popular vote isn't completely irrelevant it gives an indication on how popular candidates are overall and when you start getting into somewhat large leads it shows one candidate is going to struggle more than the other. Emerson's poll is currently at the lower end of recent polling in PA with the higher end around 8 or 9. So currently it will be a slight concern but unless a trend develops PA still looking like falling Biden's way based purely on polls.
On the one hand: Scott finding pertinent, relevant tweets and posting them here where many people don't have the time to find them directly.
On the other: paragraphs and paragraphs of people saying how they could never support Boris but they support Boris and how they want him gone but want him to stay until he has done the thing they don't support him for.
I'll take the former.
This may be a complete non sequitur, but I value both Scott and Big_G.
Having Big_G and HYFUD on the site - a former and current prominent local Tory party activist - provides a fascinating real time picture of the evolution of the Conservative Party. It feels like a case study of the application of Gresham's Law to politics.
One reason that this site is so good is the variety of political viewpoints represented, and by and large the civil nature of debate. That really is something quite unusual on the Internet and social media.
I quite like some of @Scott_P tweets, but his much rarer original author posts are very good indeed.
A bit of cooling off time is not such a bad thing though, hope @malcolmg has had his 3 match ban over soon.
As if by magic I have been pardoned, I will be on my best behaviour.
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
No it isn't. It's full of the same conjecture and bubble-thinking that led to so many people convincing themselves that Corbyn was going to march to victory.
I believe that the actual, full poll is 51% Biden.
The 21% lead stuff comes from cherry picking the data from after the Trump Covid diagnosis - it was a five day poll.
I haven't seen more detail yet, but that sounds close to people taking un-adjusted sub-samples from raw polling data, to me.
You're probably right in that.
The Boston Herald says: "In two days of polling before Trump got COVID, the president trailed Biden by just a 46%-41% margin. In the three days of polling after the coronavirus diagnosis, Biden held a 55%-34% lead. That means Biden’s lead grew by a whopping 16 points from pre-COVID to post-COVID. Among all the 1,003 registered, likely voters in the nationwide Franklin Pierce-Herald poll, Biden now holds a 51%-37% lead over Trump less than a month before Election Day."
However, the splitting of polling data shouldn't be dismissed out of hand. In the distant past. I have seen Anthony Wells run threads with heavy caveats that breaking down YouGov polling when it spans a major event, so it's not entirely spurious when analysed carefully (unlike the way the Boston Herald reported it mind). The sub-samples won't be properly weighted so there will be an extra element of randomness in there. The 16% difference here is so marked that you shouldn't disregard the strong likelihood that there was some change, even if it may have been small. The fact that the samples are reasonably balanced (40% and 60%) before and after Trump fessed up to having Covid also helps.
So I think it's reasonable to conclude that a poll showing Biden with a 14% lead contains compelling evidence that the lead would most likely have been a bit larger if the poll had been conducted to the same methodology with the same full sample to a slightly later timeframe.
I believe that the actual, full poll is 51% Biden.
The 21% lead stuff comes from cherry picking the data from after the Trump Covid diagnosis - it was a five day poll.
I haven't seen more detail yet, but that sounds close to people taking un-adjusted sub-samples from raw polling data, to me.
You're probably right in that.
The Boston Herald says: "In two days of polling before Trump got COVID, the president trailed Biden by just a 46%-41% margin. In the three days of polling after the coronavirus diagnosis, Biden held a 55%-34% lead. That means Biden’s lead grew by a whopping 16 points from pre-COVID to post-COVID. Among all the 1,003 registered, likely voters in the nationwide Franklin Pierce-Herald poll, Biden now holds a 51%-37% lead over Trump less than a month before Election Day."
However, the splitting of polling data shouldn't be dismissed out of hand. In the distant past. I have seen Anthony Wells run threads with heavy caveats that breaking down YouGov polling when it spans a major event, so it's not entirely spurious when analysed carefully (unlike the way the Boston Herald reported it mind). The sub-samples won't be properly weighted so there will be an extra element of randomness in there. The 16% difference here is so marked that you shouldn't disregard the strong likelihood that there was some change, even if it may have been small. The fact that the samples are reasonably balanced (40% and 60%) before and after Trump fessed up to having Covid also helps.
So I think it's reasonable to conclude that a poll showing Biden with a 14% lead contains compelling evidence that the lead would most likely have been a bit larger if the poll had been conducted to the same methodology with the same full sample to a slightly later timeframe.
Saying there are indication that the lead might have been bigger if the polling was all after the event is one thing.
Putting up a headline say 21% is going too far, I think.
As 2016 proved the national popular vote is irrelevant anyway, it is key swing states like PA that matter and Biden now down to a 4% lead there with Emerson today
BTW and O/T @malcolmg sends his regards. He is apparently banned at the moment, not sure why.
Personally I miss his colourful contributions.
No! Malcy may be obnoxious but he`s OUR obnoxious.
Bring him back.
Thank you Stocky and hey presto I am back.
Hello again Malcolm. Didn't realize you'd been banned. I assumed you were on a break. Jealous now. Wish I could be banned. It spells kudos and street cred. Like getting the cane at school. Come out of the Head's study, cheeks red - face too - but not crying. Cheers ring out across the playground. You grin and salute. Bit of a hero. I speak from observation rather than experience sadly. I was never that boy. It was always someone else. Plus ca change.
Thanks kinabalu, you will be surprised to hear I got the belt a few times, hand red as well as face for sure. We had some sadistic teachers in olden times.
I believe that the actual, full poll is 51% Biden.
The 21% lead stuff comes from cherry picking the data from after the Trump Covid diagnosis - it was a five day poll.
I haven't seen more detail yet, but that sounds close to people taking un-adjusted sub-samples from raw polling data, to me.
You're probably right in that.
The Boston Herald says: "In two days of polling before Trump got COVID, the president trailed Biden by just a 46%-41% margin. In the three days of polling after the coronavirus diagnosis, Biden held a 55%-34% lead. That means Biden’s lead grew by a whopping 16 points from pre-COVID to post-COVID. Among all the 1,003 registered, likely voters in the nationwide Franklin Pierce-Herald poll, Biden now holds a 51%-37% lead over Trump less than a month before Election Day."
However, the splitting of polling data shouldn't be dismissed out of hand. In the distant past. I have seen Anthony Wells run threads with heavy caveats that breaking down YouGov polling when it spans a major event, so it's not entirely spurious when analysed carefully (unlike the way the Boston Herald reported it mind). The sub-samples won't be properly weighted so there will be an extra element of randomness in there. The 16% difference here is so marked that you shouldn't disregard the strong likelihood that there was some change, even if it may have been small. The fact that the samples are reasonably balanced (40% and 60%) before and after Trump fessed up to having Covid also helps.
So I think it's reasonable to conclude that a poll showing Biden with a 14% lead contains compelling evidence that the lead would most likely have been a bit larger if the poll had been conducted to the same methodology with the same full sample to a slightly later timeframe.
I believe that the actual, full poll is 51% Biden.
The 21% lead stuff comes from cherry picking the data from after the Trump Covid diagnosis - it was a five day poll.
I haven't seen more detail yet, but that sounds close to people taking un-adjusted sub-samples from raw polling data, to me.
You're probably right in that.
The Boston Herald says: "In two days of polling before Trump got COVID, the president trailed Biden by just a 46%-41% margin. In the three days of polling after the coronavirus diagnosis, Biden held a 55%-34% lead. That means Biden’s lead grew by a whopping 16 points from pre-COVID to post-COVID. Among all the 1,003 registered, likely voters in the nationwide Franklin Pierce-Herald poll, Biden now holds a 51%-37% lead over Trump less than a month before Election Day."
However, the splitting of polling data shouldn't be dismissed out of hand. In the distant past. I have seen Anthony Wells run threads with heavy caveats that breaking down YouGov polling when it spans a major event, so it's not entirely spurious when analysed carefully (unlike the way the Boston Herald reported it mind). The sub-samples won't be properly weighted so there will be an extra element of randomness in there. The 16% difference here is so marked that you shouldn't disregard the strong likelihood that there was some change, even if it may have been small. The fact that the samples are reasonably balanced (40% and 60%) before and after Trump fessed up to having Covid also helps.
So I think it's reasonable to conclude that a poll showing Biden with a 14% lead contains compelling evidence that the lead would most likely have been a bit larger if the poll had been conducted to the same methodology with the same full sample to a slightly later timeframe.
Saying there are indication that the lead might have been bigger if the polling was all after the event is one thing.
Putting up a headline say 21% is going too far, I think.
As 2016 proved the national popular vote is irrelevant anyway, it is key swing states like PA that matter and Biden now down to a 4% lead there with Emerson today
Actually, that's quite alarming. I understood, having raised the point and been re-assured, that if the app showed positive, then one would be contacted (fairly) quickly to establish roughly at what time the positive contact took place, especially if like Mr Waugh, one has been out and about. If, in effect, the warning is being triggered by being (let us say) back to back with someone in a pub when in fact the chance of interchange of virus is low, then as, Mr S suggests, the app is rapidly going to be come unpopular and discontinued. Or, as well as 'you need to self-isolate' it should authorise a test as soon as as.
There are two different issues here. The alert itself is quick, it will trigger when the app finds a matching ID which has a self-reported positive test result. I don't know how frequently the app checks, but I would guess at several times a day. But the time between you being near to someone and them reporting a positive test result is likely to be many days, as most people will have been infectious before getting symptoms and then tested, or having been themselves exposed and told to get a test by health authorities.
Maybe F1 should follow the example of cricket and go for a quick-fire, slimmed-down, crowd-pleasing variant for the benefit for spectators too "tired" to concentrate? Circular track, 10 laps, 80mph max.
F1 on an oval would be a terrific spectacle. eg a night race at Chicagoland speedway. I don't think the tyres could take it though. They definitely should ditch one of the legacy tracks like Spa or Brands Hatch and replace it with a roval course like Charlotte.
Excellent news for my friends in Greece. Despite the virus things are not so bad everywhere in Greece now, compared to a couple of years ago.
Greece has an excellent centre right government now too
Surely you have a poll to back that statement up, why is it excellent? Because it’s centre right?
Greece has dealt well with Covid, so far. They locked down fast and hard in March, and succeeded in quickly suppressing the virus. This meant that they were in a position to receive a decent number of tourists over the summer. Of course, they've paid a price for this in the form of a larger second wave, but it looks like they're slowly getting to grips with that too. The contrast with Spain couldn't be starker.
This is the problem, isn't it? The app needs to be sensitive (in the sense of triggering for the vast majority of people who are in fact infected due to close contact) to be effective in limiting spread, but that likely implies - unless the algorithms discriminate spectacularly well - a low specificity (lots of triggers for people who have not in fact been infected). As people start to notice and report this on Twitter and the like, support for the app and use of it will likely fall, even if it is doing a good job at reducing infections overall.
To maintain support, there needs to be some analysis - and publicity - of how much the app is helping to reduce infections.
Actually, that's quite alarming. I understood, having raised the point and been re-assured, that if the app showed positive, then one would be contacted (fairly) quickly to establish roughly at what time the positive contact took place, especially if like Mr Waugh, one has been out and about. If, in effect, the warning is being triggered by being (let us say) back to back with someone in a pub when in fact the chance of interchange of virus is low, then as, Mr S suggests, the app is rapidly going to be come unpopular and discontinued. Or, as well as 'you need to self-isolate' it should authorise a test as soon as as.
How would the government be able to contact you? The system Apple and Google put in place makes it decentralised and anonymous.
This is one reason why initally the government didn't want to use the Apple / Google API, because by doing so you start with one arm tied behind your back.
The government can't win on this. The public won't accept a South Korean system, which spies on your every move, interaction, transaction, and the big tech companies have decided for the world that to run on bluetooth tech on their operating systems they have decided the rules of the game must be suboptimal for actually tracking and tracing people.
I believe that the actual, full poll is 51% Biden.
The 21% lead stuff comes from cherry picking the data from after the Trump Covid diagnosis - it was a five day poll.
I haven't seen more detail yet, but that sounds close to people taking un-adjusted sub-samples from raw polling data, to me.
You're probably right in that.
The Boston Herald says: "In two days of polling before Trump got COVID, the president trailed Biden by just a 46%-41% margin. In the three days of polling after the coronavirus diagnosis, Biden held a 55%-34% lead. That means Biden’s lead grew by a whopping 16 points from pre-COVID to post-COVID. Among all the 1,003 registered, likely voters in the nationwide Franklin Pierce-Herald poll, Biden now holds a 51%-37% lead over Trump less than a month before Election Day."
However, the splitting of polling data shouldn't be dismissed out of hand. In the distant past. I have seen Anthony Wells run threads with heavy caveats that breaking down YouGov polling when it spans a major event, so it's not entirely spurious when analysed carefully (unlike the way the Boston Herald reported it mind). The sub-samples won't be properly weighted so there will be an extra element of randomness in there. The 16% difference here is so marked that you shouldn't disregard the strong likelihood that there was some change, even if it may have been small. The fact that the samples are reasonably balanced (40% and 60%) before and after Trump fessed up to having Covid also helps.
So I think it's reasonable to conclude that a poll showing Biden with a 14% lead contains compelling evidence that the lead would most likely have been a bit larger if the poll had been conducted to the same methodology with the same full sample to a slightly later timeframe.
I believe that the actual, full poll is 51% Biden.
The 21% lead stuff comes from cherry picking the data from after the Trump Covid diagnosis - it was a five day poll.
I haven't seen more detail yet, but that sounds close to people taking un-adjusted sub-samples from raw polling data, to me.
You're probably right in that.
The Boston Herald says: "In two days of polling before Trump got COVID, the president trailed Biden by just a 46%-41% margin. In the three days of polling after the coronavirus diagnosis, Biden held a 55%-34% lead. That means Biden’s lead grew by a whopping 16 points from pre-COVID to post-COVID. Among all the 1,003 registered, likely voters in the nationwide Franklin Pierce-Herald poll, Biden now holds a 51%-37% lead over Trump less than a month before Election Day."
However, the splitting of polling data shouldn't be dismissed out of hand. In the distant past. I have seen Anthony Wells run threads with heavy caveats that breaking down YouGov polling when it spans a major event, so it's not entirely spurious when analysed carefully (unlike the way the Boston Herald reported it mind). The sub-samples won't be properly weighted so there will be an extra element of randomness in there. The 16% difference here is so marked that you shouldn't disregard the strong likelihood that there was some change, even if it may have been small. The fact that the samples are reasonably balanced (40% and 60%) before and after Trump fessed up to having Covid also helps.
So I think it's reasonable to conclude that a poll showing Biden with a 14% lead contains compelling evidence that the lead would most likely have been a bit larger if the poll had been conducted to the same methodology with the same full sample to a slightly later timeframe.
Saying there are indication that the lead might have been bigger if the polling was all after the event is one thing.
Putting up a headline say 21% is going too far, I think.
As 2016 proved the national popular vote is irrelevant anyway, it is key swing states like PA that matter and Biden now down to a 4% lead there with Emerson today
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
Compare and contrast 538 2016 - 2020. Not quite the same vibe.
To me these are the key charts. Clinton's support always looked soft, with wildly oscillating poll leads. Biden's lead looks like it is built on far firmer foundations. Americans were willing to give Trump a chance in 2016, frequently against their better judgement. In 2020 I think minds are firmly made up.
Consider the time period missing from that chart. It strongly suggests only Covid is beating Trump. The announcement of a vaccine would be expected to change everything.
Not necessarily. It looks to me like it shows Trump benefiting from a brief rally round the leader effect when the Covid crisis started hitting mid march to beginning April - as did most heads of government. Prior to that it shows Biden with a better chance - but that was when Sanders was still running.
Plus any decent model will have more uncertainty the further away you are from the election - so if things stay the same Biden's lead will equal a greater and greater likelihood of him winning as we get closer to the election.
Plus I'm not sure which chart you are using - it doesn't seem to be the 538 once quoted above.
If you look at approval ratings you probably get a more consistent picture over time, which doesn't depend on the Dem's nomination process, decreasing uncertainty as we get closer to the election, or temporary effect of the Covid rally round the leader. https://projects.fivethirtyeight.com/trump-approval-ratings/ Shows Trump doing about the same now as before the end-March Covid rally. If anything his approval ratings are slightly better now than before Covid.
The chart is from the Economist prediction model, because 538 doesn't show that far back.
It isn't a rally round the leader effect in March, because Covid didn't hit the US until later than Europe and the UK. It goes south for Trump after the deaths start piling up.
The point is that the situation before Covid was very different. Trump had a very good chance of winning. It's only Covid that has changed that. This is not what I would describe as a "firm foundation" for Biden - because a lot could be forgiven in the relief that a vaccine would bring.
Well, yes his chances looked better many months ago because who is going to predict with any certainty that far in advance? And what is going into those models? I suspect that sitting president + good economic data = good chance of reelection for many models, at that point. But Trump has been uniquely consistently unpopular as a president, so I don't think it's fair to say he was heading for victory until Covid hit, although some of his responses to Covid ("inject bleach") probably hurt his chances a bit.
I think the only fair comparison is with the approval ratings to see if Trump is doing better or worse before or after Covid. If anything his approval ratings are a little bit better now.
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
With all due respect to 'American Greatness', one can both believe that the American Left desperately needs to be 'chastened' (as another of their headlines has it) and that Donald Trump is simply too stupid and malevolent to be the man to do it as President.
Biden is winning precisely because he is the epitome of moderation, in the mould of Barack 'I would have been considered a moderate Republican in the 80s' Obama. The US progressives are pleased Trump's on his way out, but in tears about the re-emergence of Democratic centrism ('I beat the socialist!' - Joseph Robinette Biden) as a dominant electoral force. Ironically, Biden and Harris, who could collectively hold the Presidency for the next 12 years when she takes over in 2024, will do more to chasten the left than Trump could, despite his good instincts on some cultural issues.
Maybe F1 should follow the example of cricket and go for a quick-fire, slimmed-down, crowd-pleasing variant for the benefit for spectators too "tired" to concentrate? Circular track, 10 laps, 80mph max.
F1 on an oval would be a terrific spectacle. eg a night race at Chicagoland speedway. I don't think the tyres could take it though. They definitely should ditch one of the legacy tracks like Spa or Brands Hatch and replace it with a roval course like Charlotte.
I've watched a GP at Brands Hatch, but then I am very old.
This is the problem, isn't it? The app needs to be sensitive (in the sense of triggering for the vast majority of people who are in fact infected due to close contact) to be effective in limiting spread, but that likely implies - unless the algorithms discriminate spectacularly well - a low specificity (lots of triggers for people who have not in fact been infected). As people start to notice and report this on Twitter and the like, support for the app and use of it will likely fall, even if it is doing a good job at reducing infections overall.
To maintain support, there needs to be some analysis - and publicity - of how much the app is helping to reduce infections.
Actually, that's quite alarming. I understood, having raised the point and been re-assured, that if the app showed positive, then one would be contacted (fairly) quickly to establish roughly at what time the positive contact took place, especially if like Mr Waugh, one has been out and about. If, in effect, the warning is being triggered by being (let us say) back to back with someone in a pub when in fact the chance of interchange of virus is low, then as, Mr S suggests, the app is rapidly going to be come unpopular and discontinued. Or, as well as 'you need to self-isolate' it should authorise a test as soon as as.
How would the government be able to contact you? The system Apple and Google put in place makes it decentralised and anonymous.
This is one reason why initally the government didn't want to use the Apple / Google API, because by doing so you start with one arm tied behind your back.
The government can't win on this. The public won't accept a South Korean system, which spies on your every move, interaction, transaction, and the big tech companies have decided for the world that to run on bluetooth tech on their operating systems they have decided the rules of the game must be suboptimal for actually tracking and tracing people.
See the point, but if the app can tell you ought to self-isolate, how about a warning on the app that you should get a test? Plus some sort of code to indicate a genuine request?
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
The arguments I’ve seen boil down to polls can’t be wrong twice.
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
The arguments I’ve seen boil down to polls can’t be wrong twice.
... unlikely to be wrong in the same direction twice. Weak, but stronger than your version.
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
Polls in 2016 didn't weight by education which meant they were undersampling High School or lower which is a key Trump demographic. That's been fixed so the same mistake shouldn't happen again. We aren't comparing like for like.
Clinton's share wasn't overestimated. There were simply a lot of undecideds that broke for Trump last time.
Biden's share is much higher. Over 50% in key states like PA quite frequently and over 50% nationally. Clinton didn't achieve this.
There are much fewer undecideds this time for Trump to gain. Even if he gains the undecideds again, he simply can't catch up with Biden if Biden's own share is already over 50%.
Clinton's lead was soft. It kept dropping and picking back up as many voters were uncomfortable with both Trump and Clinton. Biden's lead has been absolutely rock solid.
So the polling error should have been fixed and an even bigger polling error would need to happen this time than last time.
Excellent news for my friends in Greece. Despite the virus things are not so bad everywhere in Greece now, compared to a couple of years ago.
Greece has an excellent centre right government now too
Surely you have a poll to back that statement up, why is it excellent? Because it’s centre right?
Greece has dealt well with Covid, so far. They locked down fast and hard in March, and succeeded in quickly suppressing the virus. This meant that they were in a position to receive a decent number of tourists over the summer. Of course, they've paid a price for this in the form of a larger second wave, but it looks like they're slowly getting to grips with that too. The contrast with Spain couldn't be starker.
We won’t know Spain’s current totals till Friday when they add back the confirmed cases, was shot down here three weeks ago for pointing out that the UK was on the same course as Spain.
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
With all due respect to 'American Greatness', one can both believe that the American Left desperately needs to be 'chastened' (as another of their headlines has it) and that Donald Trump is simply too stupid and malevolent to be the man to do it as President.
Biden is winning precisely because he is the epitome of moderation, in the mould of Barack 'I would have been considered a moderate Republican in the 80s' Obama. The US progressives are pleased Trump's on his way out, but in tears about the re-emergence of Democratic centrism ('I beat the socialist!' - Joseph Robinette Biden) as a dominant electoral force. Ironically, Biden and Harris, who could collectively hold the Presidency for the next 12 years when she takes over in 2024, will do more to chasten the left than Trump could, despite his good instincts on some cultural issues.
Funnily enough, I share quite a bit of that view. Ironically, having Biden in as President would take away some of the power of the left in the Democrats. Worth remembering many of the House seats won in 2018 were not campaign on the basis of opposition to Trump but on issues such as Healthcare. The left in the Democrats need a Trump like figure to thrive.
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
The arguments I’ve seen boil down to polls can’t be wrong twice.
The simple answer is that:
1. Biden's worst position in the polls sees him President whereas Clinton's support waxed and waned, with Trump leading as late as early October. Biden could lose this late peak and still win.
2. Methodologies - particularly academic qualification crossbreaks - have been updated.
3. Biden's polling is also more stable.
The only contrary evidence, as HYFUD points out, is that Biden's marginal state polling has not improved as much as his national polling implies (but would still make him President)
This is the problem, isn't it? The app needs to be sensitive (in the sense of triggering for the vast majority of people who are in fact infected due to close contact) to be effective in limiting spread, but that likely implies - unless the algorithms discriminate spectacularly well - a low specificity (lots of triggers for people who have not in fact been infected). As people start to notice and report this on Twitter and the like, support for the app and use of it will likely fall, even if it is doing a good job at reducing infections overall.
To maintain support, there needs to be some analysis - and publicity - of how much the app is helping to reduce infections.
Actually, that's quite alarming. I understood, having raised the point and been re-assured, that if the app showed positive, then one would be contacted (fairly) quickly to establish roughly at what time the positive contact took place, especially if like Mr Waugh, one has been out and about. If, in effect, the warning is being triggered by being (let us say) back to back with someone in a pub when in fact the chance of interchange of virus is low, then as, Mr S suggests, the app is rapidly going to be come unpopular and discontinued. Or, as well as 'you need to self-isolate' it should authorise a test as soon as as.
As someone else posted, the delay is likely due to the other person only just getting test (or the results) and uploading the result - not sure how many days before the positive people who were in contact will be notified, probably the notification system is being constantly tweaked based on findings (at least, I hope so).
Real problem of alarm fatigue though. I've done some work on NEWS in the past (National Early Warning Score) use in hospitals to detect deteriorating patients from vital signs. It's statistically a very good system (despite the RCP doing some interesting things with triggering levels) with area under the ROC curve, measuring ability to discriminate those in danger from those not in danger, of > 0.8, but our analysis suggested that probably only 7 out of 8 alerts from the system were false alarms or would appear so to staff (as nothing bad subsequently happened). I'd be very surprised if the algorithms for alerting in the app get any where near that good, particularly given problems like OKC points out of close proximity but facing different directions, which may have very low risk in a well ventilated or outside setting.
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
No it isn't. It's full of the same conjecture and bubble-thinking that led to so many people convincing themselves that Corbyn was going to march to victory.
Let's see. I've always said more attention should be paid to what are people are doing on the ground rather than what they are saying. I'm sure you will enlighten us all though with your ground breaking analysis.
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
Number of undecideds (far lower this time round). Consistency of Biden's lead vs. all over the placedness of Clinton's.
Plus the US public has had four years of the poop flinger in the White House, and re-electing him doesn't have quite the same 'let's roll the dice' vibe as the first time round.
I would add that the pollsters have learned from their mistakes last time, but they've probably found a whole new set to make.
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
The arguments I’ve seen boil down to polls can’t be wrong twice.
You're largely correct, but the polls last time did oscillate more, giving a range of Clinton leads from plus 15 to minus 2. They tightened markedly (perhaps because of the debates?) to the range +7 to -2 from around October 20:
The outcome, of course, was a 2% lead which proved inadequate in the Electoral College. Biden's lead has been more consistently in the 8-10 range and has recently tended to widen. If this is still the position in two weeks' time, I think it will be much more certain, but at present we can only really say that Biden is the favourite, both because the lead is larger and because the freak bias of the Electoral College might not repeat itself if the specific swing states fall differently. I'd see a 65-70% probability as reasonable at this point.
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
Number of undecideds (far lower this time round). Consistency of Biden's lead vs. all over the placedness of Clinton's.
Plus the US public has had four years of the poop flinger in the White House, and re-electing him doesn't have quite the same 'let's roll the dice' vibe as the first time round.
I would add that the pollsters have learned from their mistakes last time, but they've probably found a whole new set to make.
Bernie Madoff was consistently offering investors 6% return (was it 6% or 8%?). Didn't mean it was legitimate.
"I would add that the pollsters have learned from their mistakes last time, but they've probably found a whole new set to make"
Not sure that is true. Haven't people quoted on here ad infinitum that the polls in 2016 were essentially correct? If that is the case, why should they be correcting for mistakes if they were "right" in 2016 (NYT has essentially said this).
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
Let's consider national polls, because it's simpler.
The final national polls in 2016 gave Clinton a lead of 3.9%, by 45.7 - 41.8. Ultimately she won the national vote by 2.09%, by 48.18 - 46.09. So the polls were wrong, but not by much.
More importantly this lead was not enough to win the Electoral College. Clinton lost Wisconsin by 0.77% - assuming uniform national swing she would have needed a national vote lead of 2.87% to win the Presidency.
Biden's current lead in the national polls is 9.0%, by 51.4 - 42.4. Not only is Biden's lead greater, by 5.1 points, but the number of undecideds/others is down from 12.5% to 6.2%, making it a lot harder for Trump to catch up.
So the polls now are much stronger for Biden than they were in November 2016 for Clinton. That's not to say he's certain to win. They might tighten before November. They might be wrong again. It's possible for Trump to win the Electoral College with an even larger national vote deficit. Shenanigans may prevent all the votes from being cast or counted.
But as a starting point, the polls are much more strongly in Biden's favour than they were in Clinton's.
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
What you are missing is that when polls get an election systematically wrong, there is usually a polling inquest resulting in adjustments being made to the methodology that rule out a repeat. The US polling companies, at least the better ones, will have looked at why they got the 2016 result wrong particularly at state level and made some changes to their turnout model and sampling frame (etc).
It is quite possible that a flawed assumption that the 2020 presidential election will be like that in 2016 could cause Biden's support to be understated. Certainly based on the recent evidence in this country - in the 2010, 2015, 2017 and 2019 general elections, the adjustments are overdone such that any polling error in the subsequent election is more likely to be in the opposite direction to that in the last.
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
What you are missing is that when polls get an election systematically wrong, there is usually a polling inquest resulting in adjustments being made to the methodology that rule out a repeat. The US polling companies, at least the better ones, will have looked at why they got the 2016 result wrong particularly at state level and made some changes to their turnout model.
It is quite possible that a flawed assumption that the 2020 presidential election will be like that in 2016 could cause Biden's support to be understated. Certainly based on the recent evidence in this country - in the 2010, 2015, 2017 and 2019 general elections, the adjustments are overdone such that any polling error in the subsequent election is more likely to be in the opposite direction to that in the last.
But they didn't or at least the polling companies said they didn't get the 2016 election wrong. The report from the American polling organisations was that, at the national level, they got it broadly right. In which case, why would they have corrected any errors if they thought they got it right in the first place?
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
With all due respect to 'American Greatness', one can both believe that the American Left desperately needs to be 'chastened' (as another of their headlines has it) and that Donald Trump is simply too stupid and malevolent to be the man to do it as President.
Biden is winning precisely because he is the epitome of moderation, in the mould of Barack 'I would have been considered a moderate Republican in the 80s' Obama. The US progressives are pleased Trump's on his way out, but in tears about the re-emergence of Democratic centrism ('I beat the socialist!' - Joseph Robinette Biden) as a dominant electoral force. Ironically, Biden and Harris, who could collectively hold the Presidency for the next 12 years when she takes over in 2024, will do more to chasten the left than Trump could, despite his good instincts on some cultural issues.
And conversely are running on the most radical platform in ages.
Sanders (in particular) or Warren would have struggled to get through their agendas even if they'd scraped a win against Trump; Biden might just deliver something.
Whether this represents a temporary truce in order to defeat Trump, or if it's the beginning of a longer term accommodation is an interesting question.
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
Number of undecideds (far lower this time round). Consistency of Biden's lead vs. all over the placedness of Clinton's.
Plus the US public has had four years of the poop flinger in the White House, and re-electing him doesn't have quite the same 'let's roll the dice' vibe as the first time round.
I would add that the pollsters have learned from their mistakes last time, but they've probably found a whole new set to make.
Bernie Madoff was consistently offering investors 6% return (was it 6% or 8%?). Didn't mean it was legitimate.
"I would add that the pollsters have learned from their mistakes last time, but they've probably found a whole new set to make"
Not sure that is true. Haven't people quoted on here ad infinitum that the polls in 2016 were essentially correct? If that is the case, why should they be correcting for mistakes if they were "right" in 2016 (NYT has essentially said this).
The national polls were pretty good, the mid-west state-level polling not so much. The latter seem to have corrected what they think they did wrong (although of course they may be wrong about what they did wrong, or they may now be doing a different thing wrong.
Stepping aside from the polls, last time Trump had a much mocked retail offer that actually struck at what a lot of middle America were feeling. The fact it was mostly bluster we will put to one side. And his opponent effectively campaigned on nothing wrong there, la la la la.
This time, what's his offer? Keep America Great? Well no Covid and the economy still going well he might have stood a chance, but in the post covid world that doesn't work.
By this stage in 2016, he had MAGA, build the wall, drain the swamp, lock her up, ringing out every night across the news networks and social media.
This time something something Biden is old, something Harris is a leftie, none of this is really working. It feels totally different from 2016.
FYI, one thing to look out for tonight will be the audience numbers for the VP debate. The first debate was down 13% from 2016. If there is a similar fall this time round, that suggests a general trend (cord-cutting, fewer people watching TV). However, if the VP TV ratings numbers show a significantly better trend than the first Presidential debate, then that suggests voters are thinking more about the two VP candidates as possible Presidents.
Or far more people have already made up their minds this time so aren't bothering with the debates.
Compare and contrast 538 2016 - 2020. Not quite the same vibe.
To me these are the key charts. Clinton's support always looked soft, with wildly oscillating poll leads. Biden's lead looks like it is built on far firmer foundations. Americans were willing to give Trump a chance in 2016, frequently against their better judgement. In 2020 I think minds are firmly made up.
Consider the time period missing from that chart. It strongly suggests only Covid is beating Trump. The announcement of a vaccine would be expected to change everything.
Not necessarily. It looks to me like it shows Trump benefiting from a brief rally round the leader effect when the Covid crisis started hitting mid march to beginning April - as did most heads of government. Prior to that it shows Biden with a better chance - but that was when Sanders was still running.
Plus any decent model will have more uncertainty the further away you are from the election - so if things stay the same Biden's lead will equal a greater and greater likelihood of him winning as we get closer to the election.
Plus I'm not sure which chart you are using - it doesn't seem to be the 538 once quoted above.
If you look at approval ratings you probably get a more consistent picture over time, which doesn't depend on the Dem's nomination process, decreasing uncertainty as we get closer to the election, or temporary effect of the Covid rally round the leader. https://projects.fivethirtyeight.com/trump-approval-ratings/ Shows Trump doing about the same now as before the end-March Covid rally. If anything his approval ratings are slightly better now than before Covid.
The chart is from the Economist prediction model, because 538 doesn't show that far back.
It isn't a rally round the leader effect in March, because Covid didn't hit the US until later than Europe and the UK. It goes south for Trump after the deaths start piling up.
The point is that the situation before Covid was very different. Trump had a very good chance of winning. It's only Covid that has changed that. This is not what I would describe as a "firm foundation" for Biden - because a lot could be forgiven in the relief that a vaccine would bring.
Leaders who are perceived to have handled Covid competently have not seen their ratings plummet. I would suggest that it's Trump's response to Covid that has done for him rather than Covid per se
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
Polls in 2016 didn't weight by education which meant they were undersampling High School or lower which is a key Trump demographic. That's been fixed so the same mistake shouldn't happen again. We aren't comparing like for like.
Clinton's share wasn't overestimated. There were simply a lot of undecideds that broke for Trump last time.
Biden's share is much higher. Over 50% in key states like PA quite frequently and over 50% nationally. Clinton didn't achieve this.
There are much fewer undecideds this time for Trump to gain. Even if he gains the undecideds again, he simply can't catch up with Biden if Biden's own share is already over 50%.
Clinton's lead was soft. It kept dropping and picking back up as many voters were uncomfortable with both Trump and Clinton. Biden's lead has been absolutely rock solid.
So the polling error should have been fixed and an even bigger polling error would need to happen this time than last time.
I'd also add the recent changes to polling to pick up weighting by education was tested in the 2018 midterms and did a pretty good job with those results, so the general consensus is what was missed in 2016 is now being picked up in 2020. of course if another demographic is being missed out we dont know but nothing to suggest there is currently
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
What you are missing is that when polls get an election systematically wrong, there is usually a polling inquest resulting in adjustments being made to the methodology that rule out a repeat. The US polling companies, at least the better ones, will have looked at why they got the 2016 result wrong particularly at state level and made some changes to their turnout model.
It is quite possible that a flawed assumption that the 2020 presidential election will be like that in 2016 could cause Biden's support to be understated. Certainly based on the recent evidence in this country - in the 2010, 2015, 2017 and 2019 general elections, the adjustments are overdone such that any polling error in the subsequent election is more likely to be in the opposite direction to that in the last.
But they didn't or at least the polling companies said they didn't get the 2016 election wrong. The report from the American polling organisations was that, at the national level, they got it broadly right. In which case, why would they have corrected any errors if they thought they got it right in the first place?
No, they overstated Clinton's margin by 1% at the national level and by much more at the state level, and will be particularly concerned about the latter. They know that, whatever they say for public consumption, and to defend their reputations to clients. TBH I think you're being naiive. They know they're judged by results and will be doing their utmost not to repeat the errors of 2016.
Certainly based on the recent evidence in this country - in the 2010, 2015, 2017 and 2019 general elections, the adjustments are overdone such that any polling error in the subsequent election is more likely to be in the opposite direction to that in the last.
The problem with this sort of argument is that polling companies are not machines that are bound to repeat the errors of the past. The humans who work in them will also notice this cute pattern and may decide it's something they have to avoid in future.
So for the next GE they may make a particular effort not to overdo their adjustments, and the error might not oscillate, and may even grow further in the direction of 2019.
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
With all due respect to 'American Greatness', one can both believe that the American Left desperately needs to be 'chastened' (as another of their headlines has it) and that Donald Trump is simply too stupid and malevolent to be the man to do it as President.
Biden is winning precisely because he is the epitome of moderation, in the mould of Barack 'I would have been considered a moderate Republican in the 80s' Obama. The US progressives are pleased Trump's on his way out, but in tears about the re-emergence of Democratic centrism ('I beat the socialist!' - Joseph Robinette Biden) as a dominant electoral force. Ironically, Biden and Harris, who could collectively hold the Presidency for the next 12 years when she takes over in 2024, will do more to chasten the left than Trump could, despite his good instincts on some cultural issues.
Funnily enough, I share quite a bit of that view. Ironically, having Biden in as President would take away some of the power of the left in the Democrats. Worth remembering many of the House seats won in 2018 were not campaign on the basis of opposition to Trump but on issues such as Healthcare. The left in the Democrats need a Trump like figure to thrive.
Indeed and if Biden is President the GOP will likely win Congress in 2022 in the usual midterm protest and freed of Trump leaving him as a Bill Clinton type centrist figure, especially if the GOP keep the Senate too as they won in 1994 but not in 2010 despite winning the House in both midterms.
Trump's re election though would likely lead to a Democratic landslide in the 2022 midterm elections, especially as he would be likely to only have scraped re election and a Democratic Congress and a likely Democratic President in 2024
Latest Pennsylvania poll is only a 5%, gap, 50 - 45. Trafalgar which did well last time, I know it slants Republican, has things much much close. Ohio yesterday in another poll a dead heat. Queues to vote taking 3 hours in some places, expect that to be seven in polling day. People will get fed up and not go, probably hit Democrats harder. Still bet on Trump.
On the one hand: Scott finding pertinent, relevant tweets and posting them here where many people don't have the time to find them directly.
On the other: paragraphs and paragraphs of people saying how they could never support Boris but they support Boris and how they want him gone but want him to stay until he has done the thing they don't support him for.
I'll take the former.
This may be a complete non sequitur, but I value both Scott and Big_G.
Having Big_G and HYFUD on the site - a former and current prominent local Tory party activist - provides a fascinating real time picture of the evolution of the Conservative Party. It feels like a case study of the application of Gresham's Law to politics.
One reason that this site is so good is the variety of political viewpoints represented, and by and large the civil nature of debate. That really is something quite unusual on the Internet and social media.
I quite like some of @Scott_P tweets, but his much rarer original author posts are very good indeed.
A bit of cooling off time is not such a bad thing though, hope @malcolmg has had his 3 match ban over soon.
As if by magic I have been pardoned, I will be on my best behaviour.
Welcome back. I have missed my daily dose of comedy abuse Malc. Hope all is well with you and yours sir.
Stepping aside from the polls, last time Trump had a much mocked retail offer that actually struck at what a lot of middle America were feeling. The fact it was mostly bluster we will put to one side. And his opponent effectively campaigned on nothing wrong there, la la la la.
This time, what's his offer? Keep America Great? Well no Covid and the economy still going well he might have stood a chance, but in the post covid world that doesn't work.
By this stage in 2016, he had MAGA, build the wall, drain the swamp, lock her up, ringing out every night across the news networks and social media.
This time something something Biden is old, something Harris is a leftie, none of this is really working. It feels totally different from 2016.
I'm not sure the Biden campaign has a snappy slogan to sum up their campaign though. I'd say that for a campaign that wants to claim the mantle of change (as Trump, Brexit and Johnson have all done recently) they need to have a simple summary of what that change is.
Stepping aside from the polls, last time Trump had a much mocked retail offer that actually struck at what a lot of middle America were feeling. The fact it was mostly bluster we will put to one side. And his opponent effectively campaigned on nothing wrong there, la la la la.
This time, what's his offer? Keep America Great? Well no Covid and the economy still going well he might have stood a chance, but in the post covid world that doesn't work.
By this stage in 2016, he had MAGA, build the wall, drain the swamp, lock her up, ringing out every night across the news networks and social media.
This time something something Biden is old, something Harris is a leftie, none of this is really working. It feels totally different from 2016.
If you read the comments on here leading up to the 2016 election it really isn't that different. The majority view was that the Republicans had made a terrible mistake in having Trump running and he had no chance.
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
BTW and O/T @malcolmg sends his regards. He is apparently banned at the moment, not sure why.
Personally I miss his colourful contributions.
No! Malcy may be obnoxious but he`s OUR obnoxious.
Bring him back.
Thank you Stocky and hey presto I am back.
Hello again Malcolm. Didn't realize you'd been banned. I assumed you were on a break. Jealous now. Wish I could be banned. It spells kudos and street cred. Like getting the cane at school. Come out of the Head's study, cheeks red - face too - but not crying. Cheers ring out across the playground. You grin and salute. Bit of a hero. I speak from observation rather than experience sadly. I was never that boy. It was always someone else. Plus ca change.
Thanks kinabalu, you will be surprised to hear I got the belt a few times, hand red as well as face for sure. We had some sadistic teachers in olden times.
Well I am surprised. I bet they were all frame ups.
We had one terrible one. Mr Heavisides. He wore bell bottoms and that was best and his only harmless feature. Seemed far more interested in "discipline" than in imparting any knowledge. Most of the stuff he did would not be allowed now.
The vast majority of pollsters now weight for education to mitigate the problems in 2016.
Clinton’s leads were very up and down and there were many more undecided voters then and Trump now has a record to be judged on . The excuse then for some of the Trump vote was they disliked both candidates but gave him a chance to see if his behaviour might change during his presidency , it clearly hasn’t .
The view that some of Biden’s voters aren’t that enthusiastic for him personally misses the point . In say Obama v Romney you needed more enthusiasm because the latter wasn’t such a polarizing figure . Many of those voting for Biden absolutely loathe Trump and would walk over hot coals to make sure Trump is removed from the WH .
An anti vote counts just the same. A vote is a vote regardless of the motivation.
Trump seems to be painting himself as someone who has beaten Covid-19 and is again saying it's nothing to be afraid of. Isn't it the case that he still has to hit the worst time for Covid, up to 10 days after testing positive? He's on steroids which will be dampening down the body's response to the infection, he should keep quiet because a relapse now would be devastating for him.
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
Just to save me a bit of time, Ed, are they saying that a surge in gun sales is good for Trump?
Don't you always see reports of gun sales surging around American elections? Although given the deer hunting season typically starts around the same time I imagine there may be an element of confusing correlation with causation.
Brilliant. Who is Patti? Some figure from popular culture?
Patti Lupone, actress/singer/dancer.
Played Evita on Broadway in '79, and noted Trump's homage to her in the recent balcony scene.
Someone called Trump Covita after his balcony scene.
There is something quite Peronist about the whole Trump thing. Shirtless Ones = Deplorables?
I always thought that Trump resembled Mussolini in that pose he does with his head up and his jaw jutting out.
In googling to try to find pictures to post, I notice that other have compared Trump's appearance on the balcony with that of Mussolini. Apparently many Trump supporters were flattered by the comparison (unaware of what happened to Mussolini).
I believe that the actual, full poll is 51% Biden.
The 21% lead stuff comes from cherry picking the data from after the Trump Covid diagnosis - it was a five day poll.
I haven't seen more detail yet, but that sounds close to people taking un-adjusted sub-samples from raw polling data, to me.
You're probably right in that.
The Boston Herald says: "In two days of polling before Trump got COVID, the president trailed Biden by just a 46%-41% margin. In the three days of polling after the coronavirus diagnosis, Biden held a 55%-34% lead. That means Biden’s lead grew by a whopping 16 points from pre-COVID to post-COVID. Among all the 1,003 registered, likely voters in the nationwide Franklin Pierce-Herald poll, Biden now holds a 51%-37% lead over Trump less than a month before Election Day."
However, the splitting of polling data shouldn't be dismissed out of hand. In the distant past. I have seen Anthony Wells run threads with heavy caveats that breaking down YouGov polling when it spans a major event, so it's not entirely spurious when analysed carefully (unlike the way the Boston Herald reported it mind). The sub-samples won't be properly weighted so there will be an extra element of randomness in there. The 16% difference here is so marked that you shouldn't disregard the strong likelihood that there was some change, even if it may have been small. The fact that the samples are reasonably balanced (40% and 60%) before and after Trump fessed up to having Covid also helps.
So I think it's reasonable to conclude that a poll showing Biden with a 14% lead contains compelling evidence that the lead would most likely have been a bit larger if the poll had been conducted to the same methodology with the same full sample to a slightly later timeframe.
I believe that the actual, full poll is 51% Biden.
The 21% lead stuff comes from cherry picking the data from after the Trump Covid diagnosis - it was a five day poll.
I haven't seen more detail yet, but that sounds close to people taking un-adjusted sub-samples from raw polling data, to me.
You're probably right in that.
The Boston Herald says: "In two days of polling before Trump got COVID, the president trailed Biden by just a 46%-41% margin. In the three days of polling after the coronavirus diagnosis, Biden held a 55%-34% lead. That means Biden’s lead grew by a whopping 16 points from pre-COVID to post-COVID. Among all the 1,003 registered, likely voters in the nationwide Franklin Pierce-Herald poll, Biden now holds a 51%-37% lead over Trump less than a month before Election Day."
However, the splitting of polling data shouldn't be dismissed out of hand. In the distant past. I have seen Anthony Wells run threads with heavy caveats that breaking down YouGov polling when it spans a major event, so it's not entirely spurious when analysed carefully (unlike the way the Boston Herald reported it mind). The sub-samples won't be properly weighted so there will be an extra element of randomness in there. The 16% difference here is so marked that you shouldn't disregard the strong likelihood that there was some change, even if it may have been small. The fact that the samples are reasonably balanced (40% and 60%) before and after Trump fessed up to having Covid also helps.
So I think it's reasonable to conclude that a poll showing Biden with a 14% lead contains compelling evidence that the lead would most likely have been a bit larger if the poll had been conducted to the same methodology with the same full sample to a slightly later timeframe.
Saying there are indication that the lead might have been bigger if the polling was all after the event is one thing.
Putting up a headline say 21% is going too far, I think.
As 2016 proved the national popular vote is irrelevant anyway, it is key swing states like PA that matter and Biden now down to a 4% lead there with Emerson today
Besides, wasn't the polling lesson from 2016 that the average of the national polls is better than the state polling?
2.3% less than the average Biden Penn poll lead, forget national polls, PA, MI and WI are the states Biden must win to win the EC.
Trafalgar was better in the key swing states than the national poll average was nationally
There is indeed a poll out yesterday that has Biden's PA lead as 2.3% less than the 6.3% average (subject to rounding errors) - well within the margin of error. And another poll out yesterday had Biden's lead as 11% in PA. So what?
Also factually incorrect to say Biden must win PA, MI and WI to win the EC. That looks like the most straightforward path, but there are other ways he could plausibly win.
I bet at least one pollster does better this time around than Trafalgar in PA, MI and WI.
Latest Pennsylvania poll is only a 5%, gap, 50 - 45. Trafalgar which did well last time, I know it slants Republican, has things much much close. Ohio yesterday in another poll a dead heat. Queues to vote taking 3 hours in some places, expect that to be seven in polling day. People will get fed up and not go, probably hit Democrats harder. Still bet on Trump.
You should expect that to hurt Dems more, provided some astonishingly dumb person with a lot of influence with GOP voters didn't dissuade them from voting by post.
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
With all due respect to 'American Greatness', one can both believe that the American Left desperately needs to be 'chastened' (as another of their headlines has it) and that Donald Trump is simply too stupid and malevolent to be the man to do it as President.
Biden is winning precisely because he is the epitome of moderation, in the mould of Barack 'I would have been considered a moderate Republican in the 80s' Obama. The US progressives are pleased Trump's on his way out, but in tears about the re-emergence of Democratic centrism ('I beat the socialist!' - Joseph Robinette Biden) as a dominant electoral force. Ironically, Biden and Harris, who could collectively hold the Presidency for the next 12 years when she takes over in 2024, will do more to chasten the left than Trump could, despite his good instincts on some cultural issues.
Funnily enough, I share quite a bit of that view. Ironically, having Biden in as President would take away some of the power of the left in the Democrats. Worth remembering many of the House seats won in 2018 were not campaign on the basis of opposition to Trump but on issues such as Healthcare. The left in the Democrats need a Trump like figure to thrive.
The opposite is true of course: The Trump/MAGA/Tea Party movement grew out of a reaction to the (wrongly) perceived liberalness of Obama (and let's face it, the fact he wasn't white)
Stepping aside from the polls, last time Trump had a much mocked retail offer that actually struck at what a lot of middle America were feeling. The fact it was mostly bluster we will put to one side. And his opponent effectively campaigned on nothing wrong there, la la la la.
This time, what's his offer? Keep America Great? Well no Covid and the economy still going well he might have stood a chance, but in the post covid world that doesn't work.
By this stage in 2016, he had MAGA, build the wall, drain the swamp, lock her up, ringing out every night across the news networks and social media.
This time something something Biden is old, something Harris is a leftie, none of this is really working. It feels totally different from 2016.
If you read the comments on here leading up to the 2016 election it really isn't that different. The majority view was that the Republicans had made a terrible mistake in having Trump running and he had no chance.
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
Just to save me a bit of time, Ed, are they saying that a surge in gun sales is good for Trump?
Don't you always see reports of gun sales surging around American elections? Although given the deer hunting season typically starts around the same time I imagine there may be an element of confusing correlation with causation.
Deer hunting, yes of course, like De Niro! And on that topic I hope people remember my learned piece with that theme of a few months ago calling WH2020 as a Biden landslide.
Maybe this is a daft question and, if so, apologies.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
@Cyclefree Here is how I am looknig at the 2020 polling compared to 2016
Well, that's the rubn isn't it. But I think 2020 is going to be more accurate than 2016 because here is the educational breakdown from 2016 followed by 2020
2016
2020
Just look at what happened to the weighting by Education. It was 14% HS or less in 2016. It is now 31% in 2020
If you were to reweight the 2020 poll with the 2016 weights then it would be a 1 point race. If you reweight 2016 by 2020 then, wel, you get a lot more accurate poll with a Trump lead of 15,
There was a (very insightful as it turned out to be) post made on here at about this point before the 2016 election. It was also posted below a "Clinton is going to win article".
Point by point, let's look at what's changed ?
TheKitchenCabinet Posts: 701 October 2016 David, just on the article, a few counter-arguments:
1. The RCP four way average is now 5.3%, not 6.7% (apologies if you re referring to a 2-way). There seems to be a swing back to Trump in some polls: Rasmussen - and, ad infinitum, I am interested in the trend, not whether they are pro-Republican - had Clinton +7% on Monday and Trump +2% yesterday. There are signs this might also be happening in some states (NH, the latest poll has HRC at +3%). Biden's rcp national lead is 9%, minor parties are much less prominent this time round so 2 or 4 way polling won't make a difference; The recent trend if anything has been toward Biden
2. The RCP also does not have the LA Times poll. I know we have had the NYT article against it and the LA Times rebuttal of their points. I know Jack W says the NYT is fundamentally right; I am going to respectfully disagree as the LA Times point is a valid one, i.e. the weightings should even each other out. More importantly, though, this was the most accurate poll in 2012 and they have kept their methodology and so should not be discounted (RCP makes this very same point in an article). The latest Rasmussen is Biden +8, the latest (Quite old) LA Time is Biden +9 - bang on the average !
3. HRC seems to be hitting a wall at around 45% even with national polls showing a lead of +7-8%. The same seems to be happening in the battle ground states. Where she increases her lead, it seems to be because Trump's share falls, not that her share rises. That at least raises the possibility Trump can regain these votes and / or there are "shy Trumpsters". Biden seems to hit 50% in the states he needs
4. In the battleground states (NC, FL, NV, IA, OH), HRC has not pulled away despite this being "Trump['s worst week". In fact, as Nunu (I think) posted last night, postal votes from heavily-Democrat Cleveland do not look good and IA looks more like Trump. In NC, NV and FL, Trump is well within striking range of HRC. Biden's battleground polling is actually slightly worse than HRC according to 538 at this point; the polls have been broadly adjusted for education and Biden's numbers are higher and err 'steadier' than Clinton's. Nevertheless if you want a straw to cling onto in the Trump camp this is it
5. You mention the potential for upsets like TX but there are a few potential Democrat issues. In VA - which I assume is for HRC - the latest poll shows her with only a +3% lead; there have been reports of the Democrats putting extra resources into Minnesota (which has been trending more Republican at a local level) because of concerns about the White vote. Michigan/Wisconsin have not been forgotten about this time round by the Democrats
6. Finally, on the reports, there are enough reports coming out suggesting that at least some of these accusers have links to the Democrats. In a highly partisan election like this, I can only see this as meaning that the story only appeals to hard core Democrats and others will perhaps put them to one side. The bad news is all around Trump right now
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
Just to save me a bit of time, Ed, are they saying that a surge in gun sales is good for Trump?
And AR15 ammo, apparently. I guess having eleven guns rather than ten might make you more likely to vote, or something ?
Nick Bryant's piece today talks about this being a "dystopian" moment for America and it does feel that way sometimes, watching from afar. Yesterday I saw a clip of a Trump event in Florida. All these seniors MAGA'd and flagged up, driving a convoy of golf carts. And I thought, what a tragedy for JG Ballard that he is not around in 2020.
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
Just to save me a bit of time, Ed, are they saying that a surge in gun sales is good for Trump?
Don't you always see reports of gun sales surging around American elections? Although given the deer hunting season typically starts around the same time I imagine there may be an element of confusing correlation with causation.
Deer hunting, yes of course, like De Niro! And on that topic I hope people remember my learned piece with that theme of a few months ago calling WH2020 as a Biden landslide.
#aheadofthecurve
Was that about Trump playing Russian roulette by any chance ?
Starmer - can the PM set out the scientific evidence for the 10pm curfew? Shagger - yes, its because you voted for it. Stop undermining the country
Ultimately Boris is right, Labour voted through the non-oversight package and then let the government renew the same one and abstained on it. Starmer had the opportunity to force concessions but he flunked it and now oversight and votes are at the gift of the government who will never allow a vote on measures they know they will lose, see the withdrawal of the 10pm closing time debate/vote.
Starmer is turning out to be a rubbish politician.
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
Just to save me a bit of time, Ed, are they saying that a surge in gun sales is good for Trump?
And AR15 ammo, apparently. I guess having eleven guns rather than ten might make you more likely to vote, or something ?
Nick Bryant's piece today talks about this being a "dystopian" moment for America and it does feel that way sometimes, watching from afar. Yesterday I saw a clip of a Trump event in Florida. All these seniors MAGA'd and flagged up, driving a convoy of golf carts. And I thought, what a tragedy for JG Ballard that he is not around in 2020.
Perhaps incredibly - 538's simulations (smoothed) suggest the most likely electoral college outcome for Biden is >400... it looks like ~410 from eyeballing it.
FYI, might be of interest. It's not data-led and even I found one or two points a bit unconvincing. However, I think the point he makes about gun sales is a good one:
Just to save me a bit of time, Ed, are they saying that a surge in gun sales is good for Trump?
Don't you always see reports of gun sales surging around American elections? Although given the deer hunting season typically starts around the same time I imagine there may be an element of confusing correlation with causation.
Deer hunting, yes of course, like De Niro! And on that topic I hope people remember my learned piece with that theme of a few months ago calling WH2020 as a Biden landslide.
#aheadofthecurve
Was that about Trump playing Russian roulette by any chance ?
Got to be a non trivial chance Trump, lungs wrecked by Covid; facing an overwhelming defeat blows his own brains out shortly after the election results start to roll in I think ?
4. In the battleground states (NC, FL, NV, IA, OH), HRC has not pulled away despite this being "Trump['s worst week". In fact, as Nunu (I think) posted last night, postal votes from heavily-Democrat Cleveland do not look good and IA looks more like Trump. In NC, NV and FL, Trump is well within striking range of HRC. Biden's battleground polling is actually slightly worse than HRC according to 538 at this point; the polls have been broadly adjusted for education and Biden's numbers are higher and err 'steadier' than Clinton's. Nevertheless if you want a straw to cling onto in the Trump camp this is it
5. You mention the potential for upsets like TX but there are a few potential Democrat issues. In VA - which I assume is for HRC - the latest poll shows her with only a +3% lead; there have been reports of the Democrats putting extra resources into Minnesota (which has been trending more Republican at a local level) because of concerns about the White vote. Michigan/Wisconsin have not been forgotten about this time round by the Democrats
6. Finally, on the reports, there are enough reports coming out suggesting that at least some of these accusers have links to the Democrats. In a highly partisan election like this, I can only see this as meaning that the story only appeals to hard core Democrats and others will perhaps put them to one side. The bad news is all around Trump right now
<2/2>
Yes, I think we tend to forget the impact of the whole "HILLARY EMAILS !!!" thing.
*Betting post* - My tip yesterday of Biden 58-61% of vote at 20/1 has now come in to 13.5. That's borderline value I'd guess.
Strangely though Biden at >61% of vote has gone out to 41/1. I've nibbled again.
It's a longshot for sure, but it's not inconceivable if Trump supporters don't turnout, or postal voting makes it much easier for Dems to motivate their voters to send ballots in.
Comments
It is fun in the sim though.
It isn't a rally round the leader effect in March, because Covid didn't hit the US until later than Europe and the UK. It goes south for Trump after the deaths start piling up.
The point is that the situation before Covid was very different. Trump had a very good chance of winning. It's only Covid that has changed that. This is not what I would describe as a "firm foundation" for Biden - because a lot could be forgiven in the relief that a vaccine would bring.
If, in effect, the warning is being triggered by being (let us say) back to back with someone in a pub when in fact the chance of interchange of virus is low, then as, Mr S suggests, the app is rapidly going to be come unpopular and discontinued.
Or, as well as 'you need to self-isolate' it should authorise a test as soon as as.
https://en.wikipedia.org/wiki/Opinion_polling_for_the_next_Greek_legislative_election
https://amgreatness.com/2020/10/05/why-doesnt-it-feel-like-biden-is-winning/
https://projects.fivethirtyeight.com/polls/president-general/pennsylvania/
Which would also be consistent with Penn's Republican lean being very much the same as 2016, given Biden's 9.0% national lead
https://projects.fivethirtyeight.com/polls/president-general/national/
Besides, wasn't the polling lesson from 2016 that the average of the national polls is better than the state polling?
Emerson's poll is currently at the lower end of recent polling in PA with the higher end around 8 or 9. So currently it will be a slight concern but unless a trend develops PA still looking like falling Biden's way based purely on polls.
Welcome back.
Trafalgar was better in the key swing states than the national poll average was nationally
https://twitter.com/garywhiter/status/1313722109475123201?s=20
Biden does not need to win Penn to be president.
This is one reason why initally the government didn't want to use the Apple / Google API, because by doing so you start with one arm tied behind your back.
The government can't win on this. The public won't accept a South Korean system, which spies on your every move, interaction, transaction, and the big tech companies have decided for the world that to run on bluetooth tech on their operating systems they have decided the rules of the game must be suboptimal for actually tracking and tracing people.
What is the probability that Biden becomes President if he loses PA?
What is the probability that Biden wins PA if he becomes President?
What figures would you give for those?
Played Evita on Broadway in '79, and noted Trump's homage to her in the recent balcony scene.
But here goes: didn’t all the polls in 2016 show Clinton winning? Now all the polls are showing Biden winning. Why should they be / are they any more right than in 2016?
I really want Trump to lose (and lose badly, ideally). But I can’t help feeling that too many - possibly including the Biden leadership team - are in danger of confusing what they would like to happen with what is happening. Trump’s appeal to people more worried about law and order than about elegant speeches at Gettysburg should not be underestimated.
What am I missing?
I think the only fair comparison is with the approval ratings to see if Trump is doing better or worse before or after Covid. If anything his approval ratings are a little bit better now.
Biden is winning precisely because he is the epitome of moderation, in the mould of Barack 'I would have been considered a moderate Republican in the 80s' Obama. The US progressives are pleased Trump's on his way out, but in tears about the re-emergence of Democratic centrism ('I beat the socialist!' - Joseph Robinette Biden) as a dominant electoral force. Ironically, Biden and Harris, who could collectively hold the Presidency for the next 12 years when she takes over in 2024, will do more to chasten the left than Trump could, despite his good instincts on some cultural issues.
The mtfr clip is from 2017.
- Polls in 2016 didn't weight by education which meant they were undersampling High School or lower which is a key Trump demographic. That's been fixed so the same mistake shouldn't happen again. We aren't comparing like for like.
- Clinton's share wasn't overestimated. There were simply a lot of undecideds that broke for Trump last time.
- Biden's share is much higher. Over 50% in key states like PA quite frequently and over 50% nationally. Clinton didn't achieve this.
- There are much fewer undecideds this time for Trump to gain. Even if he gains the undecideds again, he simply can't catch up with Biden if Biden's own share is already over 50%.
- Clinton's lead was soft. It kept dropping and picking back up as many voters were uncomfortable with both Trump and Clinton. Biden's lead has been absolutely rock solid.
So the polling error should have been fixed and an even bigger polling error would need to happen this time than last time.1. Biden's worst position in the polls sees him President whereas Clinton's support waxed and waned, with Trump leading as late as early October. Biden could lose this late peak and still win.
2. Methodologies - particularly academic qualification crossbreaks - have been updated.
3. Biden's polling is also more stable.
The only contrary evidence, as HYFUD points out, is that Biden's marginal state polling has not improved as much as his national polling implies (but would still make him President)
Real problem of alarm fatigue though. I've done some work on NEWS in the past (National Early Warning Score) use in hospitals to detect deteriorating patients from vital signs. It's statistically a very good system (despite the RCP doing some interesting things with triggering levels) with area under the ROC curve, measuring ability to discriminate those in danger from those not in danger, of > 0.8, but our analysis suggested that probably only 7 out of 8 alerts from the system were false alarms or would appear so to staff (as nothing bad subsequently happened). I'd be very surprised if the algorithms for alerting in the app get any where near that good, particularly given problems like OKC points out of close proximity but facing different directions, which may have very low risk in a well ventilated or outside setting.
Consistency of Biden's lead vs. all over the placedness of Clinton's.
Plus the US public has had four years of the poop flinger in the White House, and re-electing him doesn't have quite the same 'let's roll the dice' vibe as the first time round.
I would add that the pollsters have learned from their mistakes last time, but they've probably found a whole new set to make.
https://en.wikipedia.org/wiki/Nationwide_opinion_polling_for_the_2016_United_States_presidential_election
The outcome, of course, was a 2% lead which proved inadequate in the Electoral College. Biden's lead has been more consistently in the 8-10 range and has recently tended to widen. If this is still the position in two weeks' time, I think it will be much more certain, but at present we can only really say that Biden is the favourite, both because the lead is larger and because the freak bias of the Electoral College might not repeat itself if the specific swing states fall differently. I'd see a 65-70% probability as reasonable at this point.
"I would add that the pollsters have learned from their mistakes last time, but they've probably found a whole new set to make"
Not sure that is true. Haven't people quoted on here ad infinitum that the polls in 2016 were essentially correct? If that is the case, why should they be correcting for mistakes if they were "right" in 2016 (NYT has essentially said this).
The final national polls in 2016 gave Clinton a lead of 3.9%, by 45.7 - 41.8. Ultimately she won the national vote by 2.09%, by 48.18 - 46.09. So the polls were wrong, but not by much.
More importantly this lead was not enough to win the Electoral College. Clinton lost Wisconsin by 0.77% - assuming uniform national swing she would have needed a national vote lead of 2.87% to win the Presidency.
Biden's current lead in the national polls is 9.0%, by 51.4 - 42.4. Not only is Biden's lead greater, by 5.1 points, but the number of undecideds/others is down from 12.5% to 6.2%, making it a lot harder for Trump to catch up.
So the polls now are much stronger for Biden than they were in November 2016 for Clinton. That's not to say he's certain to win. They might tighten before November. They might be wrong again. It's possible for Trump to win the Electoral College with an even larger national vote deficit. Shenanigans may prevent all the votes from being cast or counted.
But as a starting point, the polls are much more strongly in Biden's favour than they were in Clinton's.
It is quite possible that a flawed assumption that the 2020 presidential election will be like that in 2016 could cause Biden's support to be understated. Certainly based on the recent evidence in this country - in the 2010, 2015, 2017 and 2019 general elections, the adjustments are overdone such that any polling error in the subsequent election is more likely to be in the opposite direction to that in the last.
Sanders (in particular) or Warren would have struggled to get through their agendas even if they'd scraped a win against Trump; Biden might just deliver something.
Whether this represents a temporary truce in order to defeat Trump, or if it's the beginning of a longer term accommodation is an interesting question.
This time, what's his offer? Keep America Great? Well no Covid and the economy still going well he might have stood a chance, but in the post covid world that doesn't work.
By this stage in 2016, he had MAGA, build the wall, drain the swamp, lock her up, ringing out every night across the news networks and social media.
This time something something Biden is old, something Harris is a leftie, none of this is really working. It feels totally different from 2016.
So for the next GE they may make a particular effort not to overdo their adjustments, and the error might not oscillate, and may even grow further in the direction of 2019.
Trump's re election though would likely lead to a Democratic landslide in the 2022 midterm elections, especially as he would be likely to only have scraped re election and a Democratic Congress and a likely Democratic President in 2024
https://twitter.com/fnihost/status/1313302061186985985
I guess having eleven guns rather than ten might make you more likely to vote, or something ?
We had one terrible one. Mr Heavisides. He wore bell bottoms and that was best and his only harmless feature. Seemed far more interested in "discipline" than in imparting any knowledge. Most of the stuff he did would not be allowed now.
Clinton’s leads were very up and down and there were many more undecided voters then and Trump now has a record to be judged on . The excuse then for some of the Trump vote was they disliked both candidates but gave him a chance to see if his behaviour might change during his presidency , it clearly hasn’t .
The view that some of Biden’s voters aren’t that enthusiastic for him personally misses the point . In say Obama v Romney you needed more enthusiasm because the latter wasn’t such a polarizing figure . Many of those voting for Biden absolutely loathe Trump and would walk over hot coals to make sure Trump is removed from the WH .
An anti vote counts just the same. A vote is a vote regardless of the motivation.
Isn't it the case that he still has to hit the worst time for Covid, up to 10 days after testing positive? He's on steroids which will be dampening down the body's response to the infection, he should keep quiet because a relapse now would be devastating for him.
In googling to try to find pictures to post, I notice that other have compared Trump's appearance on the balcony with that of Mussolini. Apparently many Trump supporters were flattered by the comparison (unaware of what happened to Mussolini).
Also factually incorrect to say Biden must win PA, MI and WI to win the EC. That looks like the most straightforward path, but there are other ways he could plausibly win.
I bet at least one pollster does better this time around than Trafalgar in PA, MI and WI.
#aheadofthecurve
Shagger - yes, its because you voted for it. Stop undermining the country
Here is a SurveyUSA poll from October 2016
http://www.surveyusa.com/client/PollReport.aspx?g=0e1d48a5-2698-448f-8f7a-fa2c92996543
It's a Kansas Poll. In the Poll Trump leads by 11.
Trump won the state by 20.5
That is a heinous miss.
Here is a SurveryUSA poll from August 2020
http://www.surveyusa.com/client/PollReport.aspx?g=b7efcf1a-9519-4b36-95f6-5a05b8cc7f6d
It's a Kansas Poll, Trump leads by 7
Trump will win the state by....
Well, that's the rubn isn't it. But I think 2020 is going to be more accurate than 2016 because here is the educational breakdown from 2016 followed by 2020
2016
2020
Just look at what happened to the weighting by Education. It was 14% HS or less in 2016. It is now 31% in 2020
If you were to reweight the 2020 poll with the 2016 weights then it would be a 1 point race. If you reweight 2016 by 2020 then, wel, you get a lot more accurate poll with a Trump lead of 15,
That's Labour that is!
Point by point, let's look at what's changed ?
TheKitchenCabinet Posts: 701
October 2016
David, just on the article, a few counter-arguments:
1. The RCP four way average is now 5.3%, not 6.7% (apologies if you re referring to a 2-way). There seems to be a swing back to Trump in some polls: Rasmussen - and, ad infinitum, I am interested in the trend, not whether they are pro-Republican - had Clinton +7% on Monday and Trump +2% yesterday. There are signs this might also be happening in some states (NH, the latest poll has HRC at +3%). Biden's rcp national lead is 9%, minor parties are much less prominent this time round so 2 or 4 way polling won't make a difference; The recent trend if anything has been toward Biden
2. The RCP also does not have the LA Times poll. I know we have had the NYT article against it and the LA Times rebuttal of their points. I know Jack W says the NYT is fundamentally right; I am going to respectfully disagree as the LA Times point is a valid one, i.e. the weightings should even each other out. More importantly, though, this was the most accurate poll in 2012 and they have kept their methodology and so should not be discounted (RCP makes this very same point in an article). The latest Rasmussen is Biden +8, the latest (Quite old) LA Time is Biden +9 - bang on the average !
3. HRC seems to be hitting a wall at around 45% even with national polls showing a lead of +7-8%. The same seems to be happening in the battle ground states. Where she increases her lead, it seems to be because Trump's share falls, not that her share rises. That at least raises the possibility Trump can regain these votes and / or there are "shy Trumpsters". Biden seems to hit 50% in the states he needs
<1/2>
5. You mention the potential for upsets like TX but there are a few potential Democrat issues. In VA - which I assume is for HRC - the latest poll shows her with only a +3% lead; there have been reports of the Democrats putting extra resources into Minnesota (which has been trending more Republican at a local level) because of concerns about the White vote. Michigan/Wisconsin have not been forgotten about this time round by the Democrats
6. Finally, on the reports, there are enough reports coming out suggesting that at least some of these accusers have links to the Democrats. In a highly partisan election like this, I can only see this as meaning that the story only appeals to hard core Democrats and others will perhaps put them to one side. The bad news is all around Trump right now
<2/2>
Starmer is turning out to be a rubbish politician.
They didn't get the patent, but the Nobel is deserved recognition.
Strangely though Biden at >61% of vote has gone out to 41/1. I've nibbled again.
It's a longshot for sure, but it's not inconceivable if Trump supporters don't turnout, or postal voting makes it much easier for Dems to motivate their voters to send ballots in.