|
Post by curiousliberal on Aug 15, 2020 18:00:02 GMT
While polls are of course an imperfect measure, the lengths people will go to to find a way of predicting an election that doesn’t involve simply asking people how they will vote is ridiculous. I think thats the point. Given how inaccurate polling has proven to be over the past few years I think people are scratching around for something a tad more accurate at predictions. This time its the stock market next week it might be tea-leaves. It's not been that bad and 2018 saw pollsters fix a fair few of the issues with the 2016 cycle's output. There is a tendency to judge polls more harshly when they project the wrong winner by being 2 points off the final result as opposed to projecting the eventual victor while being 10 points off.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Aug 15, 2020 18:45:59 GMT
With this postal vote fiasco, it's going to be Bush v Gore on steroids
|
|
|
Post by Daft H'a'porth A'peth A'pith on Aug 15, 2020 21:01:02 GMT
With this postal vote fiasco, it's going to be Bush v Gore on steroids
Just what all the politicians want so they can blame someone other than themselves, even though the created the system and the mess.
|
|
|
Post by timrollpickering on Aug 15, 2020 22:29:08 GMT
1872: Benjamin Brown (D-Missouri) Actually Liberal Republican, a breakaway party. Brown was elected Governor on the LR ticket in 1870 and then nominated for Veep in 1872; the Democrats opted to nominate the LR ticket as part of a fusion drive. You can add: Yes 1812: Jared Ingersoll (F-Pennsylvania) - This was another complicated election with the main opposition party ultimately backing a challenger from with the incumbent's party but the Veep nominee was from their own. No 1828: John C. Calhoun (D-South Carolina) - An odd case as he was the sitting Veep but running on the challenger ticket. In the mess of the 1824 election Calhoun had picked up endorsements from both the Jackson and Adams campaign despite opposing the latter's platform and secured a majority of electoral votes. His state voted for Jackson in 1824 and they sided together as the Democrats were formed. 1820 is another odd case. Although Monroe was re-elected unopposed (bar one faithless elector), in a few states the Federalists won the electors and voted for a different Veep from Daniel Tompkins. Richard Stockton (New Jersey), Robert Goodloe Harper (Maryland) and Richard Rush (Pennsylvania) were all from states the incumbent had carried last time but Daniel Rodney (Delaware) wasn't. (Delaware would be one of the last pockets of Federalist strength into the late 1820s.)
|
|
|
Post by Devil Wincarnate on Aug 16, 2020 0:15:17 GMT
This is fun.
|
|
|
Post by London Lad on Aug 16, 2020 6:02:57 GMT
I think thats the point. Given how inaccurate polling has proven to be over the past few years I think people are scratching around for something a tad more accurate at predictions. This time its the stock market next week it might be tea-leaves. It's not been that bad and 2018 saw pollsters fix a fair few of the issues with the 2016 cycle's output. There is a tendency to judge polls more harshly when they project the wrong winner by being 2 points off the final result as opposed to projecting the eventual victor while being 10 points off. But after every bad forecast they say they have fixed their systems - until they get it wrong the next time. In 2016 - CNN had Trump with a 1% chance of winning the election, the University of Illinois had him at Zero (!) chance of winning the election and even five thirty eight only gave him a 28% chance. What sort of 'fixes' do you have to do to make that type of polling more accurate?
|
|
|
Post by London Lad on Aug 16, 2020 6:08:54 GMT
With this postal vote fiasco, it's going to be Bush v Gore on steroids But wouldn't it be fun?. The USPS is struggling now so the ability of their systems to suddenly cope with hundreds of thousands (possibly millions) of extra votes will surely be tested. It would be great though if the election went ahead and Trump just about managed to win based on a majority of 80,000 in a swing state like Florida and gets their 29 electoral votes then a week later the USPS found another 100,000 votes for Biden in a sack in the corner of one of its warehouses in Miami..
|
|
Richard Allen
Banned
Four time loser in VUKPOTY finals
Posts: 19,052
|
Post by Richard Allen on Aug 16, 2020 7:57:14 GMT
With this postal vote fiasco, it's going to be Bush v Gore on steroids But wouldn't it be fun?. The USPS is struggling now so the ability of their systems to suddenly cope with hundreds of thousands (possibly millions) of extra votes will surely be tested. It would be great though if the election went ahead and Trump just about managed to win based on a majority of 80,000 in a swing state like Florida and gets their 29 electoral votes then a week later the USPS found another 100,000 votes for Biden in a sack in the corner of one of its warehouses in Miami.. Although the thing here is that in recent years the Florida GOP have been very successful at using mail in votes while as in all things the Florida Democrats have proven useless. It has been a big factor in the GOP winning many close races.
|
|
|
Post by greenchristian on Aug 16, 2020 9:39:21 GMT
It's not been that bad and 2018 saw pollsters fix a fair few of the issues with the 2016 cycle's output. There is a tendency to judge polls more harshly when they project the wrong winner by being 2 points off the final result as opposed to projecting the eventual victor while being 10 points off. But after every bad forecast they say they have fixed their systems - until they get it wrong the next time. In 2016 - CNN had Trump with a 1% chance of winning the election, the University of Illinois had him at Zero (!) chance of winning the election and even five thirty eight only gave him a 28% chance. What sort of 'fixes' do you have to do to make that type of polling more accurate? You seem to be equating accuracy with getting the winner right, rather than getting the vote shares right. A poll that gets the winner wrong but the vote shares out by 2% is more accurate than one that gets the winner right but the vote shares out by 10$. The polls in 2016 didn't get the national vote shares that far out despite Comey dropping some potentially result-changing news too late for it to show up in polls. The problem was that only 538 saw the real potential for Trump barely squeaking out an electoral college win whilst being that far behind in the popular vote (Bush and Gore were way closer in the popular vote).
|
|
The Bishop
Labour
Down With Factionalism!
Posts: 36,607
|
Post by The Bishop on Aug 16, 2020 9:55:06 GMT
Indeed, I recall some of the 2016 campaign speculation that Clinton was more likely to be the beneficiary of a "wrong winner" result.
|
|
|
Post by manchesterman on Aug 16, 2020 10:25:18 GMT
But after every bad forecast they say they have fixed their systems - until they get it wrong the next time. In 2016 - CNN had Trump with a 1% chance of winning the election, the University of Illinois had him at Zero (!) chance of winning the election and even five thirty eight only gave him a 28% chance. What sort of 'fixes' do you have to do to make that type of polling more accurate? You seem to be equating accuracy with getting the winner right, rather than getting the vote shares right. A poll that gets the winner wrong but the vote shares out by 2% is more accurate than one that gets the winner right but the vote shares out by 10$. The polls in 2016 didn't get the national vote shares that far out despite Comey dropping some potentially result-changing news too late for it to show up in polls. The problem was that only 538 saw the real potential for Trump barely squeaking out an electoral college win whilst being that far behind in the popular vote (Bush and Gore were way closer in the popular vote).
Indeed, maybe what the pollsters should be saying is that we believe the percentage votes for each candidate will be X and Y (but vote distribution will determine who becomes POTUS). But that isnt "sexy" enough to grab attention I guess, so they then make that quantum leap to - more or less- guesswork as to how the votes will fall and that can lead to outcomes like 2016. As others have commented, the pollsters got the voting percentages about right.
|
|
|
Post by London Lad on Aug 16, 2020 11:22:26 GMT
But after every bad forecast they say they have fixed their systems - until they get it wrong the next time. In 2016 - CNN had Trump with a 1% chance of winning the election, the University of Illinois had him at Zero (!) chance of winning the election and even five thirty eight only gave him a 28% chance. What sort of 'fixes' do you have to do to make that type of polling more accurate? You seem to be equating accuracy with getting the winner right, rather than getting the vote shares right. A poll that gets the winner wrong but the vote shares out by 2% is more accurate than one that gets the winner right but the vote shares out by 10$. The polls in 2016 didn't get the national vote shares that far out despite Comey dropping some potentially result-changing news too late for it to show up in polls. The problem was that only 538 saw the real potential for Trump barely squeaking out an electoral college win whilst being that far behind in the popular vote (Bush and Gore were way closer in the popular vote).
Isn't that the whole point of polling at an election - to give us an idea of who is going to win?. Getting national share of the vote correct in an election that is decided by an Electoral College is rather pointless, especially in a country where you have the major population centre concentrated in a very few States. In the UK we dont expect polling to give us the result of a notional National Vote - we expect to get an idea of how many seats each Party will win and who will form the next Government. Why hold foreign pollsters to different standards?
|
|
|
Post by Davıd Boothroyd on Aug 16, 2020 11:24:06 GMT
Indeed, I recall some of the 2016 campaign speculation that Clinton was more likely to be the beneficiary of a "wrong winner" result. Just before the 2000 election it was thought more likely that Bush would top the poll while Gore won the electoral college. Turned out the other way round.
|
|
|
Post by greenchristian on Aug 16, 2020 11:46:47 GMT
You seem to be equating accuracy with getting the winner right, rather than getting the vote shares right. A poll that gets the winner wrong but the vote shares out by 2% is more accurate than one that gets the winner right but the vote shares out by 10$. The polls in 2016 didn't get the national vote shares that far out despite Comey dropping some potentially result-changing news too late for it to show up in polls. The problem was that only 538 saw the real potential for Trump barely squeaking out an electoral college win whilst being that far behind in the popular vote (Bush and Gore were way closer in the popular vote).
Isn't that the whole point of polling at an election - to give us an idea of who is going to win?. Getting national share of the vote correct in an election that is decided by an Electoral College is rather pointless, especially in a country where you have the major population centre concentrated in a very few States. In the UK we dont expect polling to give us the result of a notional National Vote - we expect to get an idea of how many seats each Party will win and who will form the next Government. Why hold foreign pollsters to different standards? The point of polling is to tell us what the public think. But polling is not, and cannot be, a precise science - there is always a margin of error. If the polling is showing it will be close, as was the case in 2016, then that margin of error means that there is a relatively high chance that the candidate who is polling in the lead will not win. As for UK polling, just look at the figures in the polling section. Apart from a couple of MRP models every poll gives us the result of a notional National Vote. We can try various methods to translate that into seat results, but our polling does not directly predict that.
|
|
|
Post by finsobruce on Aug 16, 2020 12:46:36 GMT
Van Buren was in league with Satan according to the lyrics.
A whole CD of these is available!
|
|
|
Post by curiousliberal on Aug 16, 2020 14:45:15 GMT
It's not been that bad and 2018 saw pollsters fix a fair few of the issues with the 2016 cycle's output. There is a tendency to judge polls more harshly when they project the wrong winner by being 2 points off the final result as opposed to projecting the eventual victor while being 10 points off. But after every bad forecast they say they have fixed their systems - until they get it wrong the next time. In 2016 - CNN had Trump with a 1% chance of winning the election, the University of Illinois had him at Zero (!) chance of winning the election and even five thirty eight only gave him a 28% chance. What sort of 'fixes' do you have to do to make that type of polling more accurate? Forecasts are different to polls and were definitely much worthier of criticism. Part of the problem was that most didn't actually take polling seriously enough or relied too much on who was ahead as opposed to the margin polls suggested they were ahead by; as you've observed, 538 was the exception to this rule, giving Trump a 28% chance. A 72% chance for HRC is Lean Clinton in my book and I don't think that was an unreasonable rating heading into election day. This is to say that the Trump victory was an upset, but that polling left room for it to be an upset within reason as opposed to a bit more of a shock (which it would have been if the election was Likely Clinton) or a total misfire for pollsters (if it had been safe Clinton). One of the key issues with polling in 2016 was the lack of proper weighting by education (when it comes to projecting turnout, among other things). Non-college educated voters tend to respond less to pollsters and because non-college educated whites trended more Republican than the national swing to Trump, the impact of their shift at the ballot box was usually underestimated in states with lots of them (see: the midwest). Similarly, non-college educated Hispanic voters were underestimated in Nevada, where Clinton outperformed polling. Pollsters and poll-based forecasts which corrected for this in 2018 tended to be much more accurate than those which just stuck to voter files from past elections.
|
|
|
Post by pragmaticidealist on Aug 16, 2020 15:56:01 GMT
Indeed, I recall some of the 2016 campaign speculation that Clinton was more likely to be the beneficiary of a "wrong winner" result. Just before the 2000 election it was thought more likely that Bush would top the poll while Gore won the electoral college. Turned out the other way round. Perhaps in part due to California being seen as somewhat competitive and Bush having Texas (which had been quite close in 1992 and 1996) as his home state. Here's a clip in which the above scenario is explicitly stated as a possibility: Also note the general tone. It really did, it seems, feel as though Gore had won.
|
|
|
Post by greatkingrat on Aug 16, 2020 17:33:34 GMT
While polls are of course an imperfect measure, the lengths people will go to to find a way of predicting an election that doesn’t involve simply asking people how they will vote is ridiculous. Isn't there some rule based on which team win a particular sports march? en.wikipedia.org/wiki/Redskins_RuleIf the Redskins win their last home match before the election, the incumbent party wins.
|
|
|
Post by manchesterman on Aug 16, 2020 19:58:36 GMT
Isn't there some rule based on which team win a particular sports march? en.wikipedia.org/wiki/Redskins_RuleIf the Redskins win their last home match before the election, the incumbent party wins. Given their recent form, trump may have a better chance relying on his current policy of massively restricting the opportunities for postal voting
|
|
|
Post by curiousliberal on Aug 17, 2020 0:02:26 GMT
Politico has a habit of playing up internal divisions for drama's sake in both blue and red teams (so I'd recommend reading with a pinch of salt), but this piece gives a good look at how the beltway-focused navel-gazing of Obama loyalists (and, more widely, Hillaryworld) underestimated Biden. I'd argue the blind spots shown here would have made up the difference between winning and losing 2016.
|
|