Looks like my "Clinton Handicap" electoral college map projection may have accurately predicted every race other than Florida. (You can find it in my video Can Trump Still Win? Posted on Oct. 30)
We'll have to wait to see how accurate it ends up being, but I do want to point out why this map was probably more right than most of the msm projections. I'll consider making a video on this, if you guys are interested, but essentially, my calculations were based on the idea that we would have a repeat of the polling bias present in 2016: that pollsters, and thus FiveThirtyEight, would be off by about as much in each state as they were then. (In fact, it looks like they were even more off.)
FiveThirtyEight and others continually insisted throughout the 2020 election that many or most pollsters had fixed their polling mistakes from 2016. They thought 'uneducated whites' was the one problem with last election's polls, so they began to weight results by education. It made sense to me that that would be a potential improvement, but I was not particularly convinced that that would do the real job.
Based in part from feed back I've received from the Trump supporters who have watched my videos and engaged with me in good faith conversations in the comments sections, I've long considered the possibility that the problem was neither the weighting of different polls (by 538 or RCP) nor the weighting of demographics within polls (done by the pollsters themselves--Gallup, Emerson, Morning Consult, IPSOS, etc).
The corruption of polls involving Trump supporters, I think, happens at the level of the respondents themselves. If you are a Trump supporter, you probably do not have kind thoughts about Universities like Monmouth or Harvard. Likewise, you probably don't like media outlets like CNN or NBC. Indeed, a tenet of Trumpism has been that the media is "the enemy of the people". So, not only should we expect that a lot of Trump supporters would not be happy to respond accurately to polls from msm outlets--we should expect that a reasonable portion might actively feel an interest to prove that the media is fake, by serving them up fake answers to a survey they know the media will tout as accurate.
Several pollsters have confused this concept with the idea of the "shy Trump supporter". Tests run by Morning Consult and others were conducted: polling by phone and online to compare results. Their tests found little-no shyness: online or by phone, there seemed to be the same number of Trump supporters. But, the issue is not "shy" support, it's active disdain.
If you want your polls to be accurate, the people you survey need to want the same thing, or at least, not really care.
The reason I think the "Clinton Handicap" dealt fairly well with this trend is the fact that the only assumption inherent to it is that we'll see an approximately steady % of Trump supporters lying to pollsters (in a random, weighted sample in a given state).
This, apparently wasn't the case in Florida, if it is indeed the cause of the polling inaccuracies we've seen since Trump entered the political arena. If it was the only cause, then Trump supporter disdain for pollsters must have increased during the past four years. (A more precise analysis of the figures in each state should give us a better understanding how far off the counts have been, and if there's desirable trends, assuming my hypothesis to be correct.)
If pollster disdain was indeed the cause of polling inaccuracies during the Trump era, and if it's coming to an end now, should we expect this issue to continue on into a Biden presidency?
Question Time
Looks like my "Clinton Handicap" electoral college map projection may have accurately predicted every race other than Florida. (You can find it in my video Can Trump Still Win? Posted on Oct. 30)
We'll have to wait to see how accurate it ends up being, but I do want to point out why this map was probably more right than most of the msm projections. I'll consider making a video on this, if you guys are interested, but essentially, my calculations were based on the idea that we would have a repeat of the polling bias present in 2016: that pollsters, and thus FiveThirtyEight, would be off by about as much in each state as they were then. (In fact, it looks like they were even more off.)
FiveThirtyEight and others continually insisted throughout the 2020 election that many or most pollsters had fixed their polling mistakes from 2016. They thought 'uneducated whites' was the one problem with last election's polls, so they began to weight results by education. It made sense to me that that would be a potential improvement, but I was not particularly convinced that that would do the real job.
Based in part from feed back I've received from the Trump supporters who have watched my videos and engaged with me in good faith conversations in the comments sections, I've long considered the possibility that the problem was neither the weighting of different polls (by 538 or RCP) nor the weighting of demographics within polls (done by the pollsters themselves--Gallup, Emerson, Morning Consult, IPSOS, etc).
The corruption of polls involving Trump supporters, I think, happens at the level of the respondents themselves. If you are a Trump supporter, you probably do not have kind thoughts about Universities like Monmouth or Harvard. Likewise, you probably don't like media outlets like CNN or NBC. Indeed, a tenet of Trumpism has been that the media is "the enemy of the people". So, not only should we expect that a lot of Trump supporters would not be happy to respond accurately to polls from msm outlets--we should expect that a reasonable portion might actively feel an interest to prove that the media is fake, by serving them up fake answers to a survey they know the media will tout as accurate.
Several pollsters have confused this concept with the idea of the "shy Trump supporter". Tests run by Morning Consult and others were conducted: polling by phone and online to compare results. Their tests found little-no shyness: online or by phone, there seemed to be the same number of Trump supporters. But, the issue is not "shy" support, it's active disdain.
If you want your polls to be accurate, the people you survey need to want the same thing, or at least, not really care.
The reason I think the "Clinton Handicap" dealt fairly well with this trend is the fact that the only assumption inherent to it is that we'll see an approximately steady % of Trump supporters lying to pollsters (in a random, weighted sample in a given state).
This, apparently wasn't the case in Florida, if it is indeed the cause of the polling inaccuracies we've seen since Trump entered the political arena. If it was the only cause, then Trump supporter disdain for pollsters must have increased during the past four years. (A more precise analysis of the figures in each state should give us a better understanding how far off the counts have been, and if there's desirable trends, assuming my hypothesis to be correct.)
If pollster disdain was indeed the cause of polling inaccuracies during the Trump era, and if it's coming to an end now, should we expect this issue to continue on into a Biden presidency?
4 years ago | [YT] | 585