The undersigned wrote both Wednesday and Friday of his concern with the declining gap between President Biden and IllegiTrump according to the most recent polls. He caveated his concern that the recent polls have not been of the highest caliber nor from the most highly-rated polling firms. But he felt beyond a reasonable doubt that if the majors conduct other polls soon they would indicate some decline also. The narrowing of the gap is insult to the soul of this Idiot Blogger, to the model of the 2020 election that he has had long constructed in his head more particularly, viz: that people have made up their minds on Trump pro and con; most, enough to win the election for Biden, have made up their minds that Trump Sucks. The model was constructed with polling data and chinked with learned commentary from non-Idiot's. The model is proved if there is little movement in the poll numbers; it is disproved by a consistent two-week decline in the gap between the two candidates, which is what we have done seed.
Comes now the estimable Charlie Cook of the eponymous Cook Political Report who on August 7 advanced the polling axiom "garbage in, garbage out." In an article headlined, "The Polling Is Imperfect, but It’s Bad News for Trump" Mr. Cook writes:
...poll aggregates and models like RealClearPolitics and FiveThirtyEight.
...there are limitations to poll aggregations. Both of the main ones include surveys that I find awfully sketchy—and a couple that I find little short of fraudulent. Nate Silver’s FiveThirtyEight, which is far more sophisticated than the more raw RealClearPolitics, attempts to address the wide range in caliber of pollsters by assigning a rating for each pollster based on their track records. That has value, though I have real questions about many of the pollster ratings. I see some pretty mediocre firms with good ratings and some of the best in the business, many political firms whose polls are 99.9 percent of the time in the private domain, with lower ratings based on the very few polls that are publicly available.
[Mixed message: 538's aggregate is "far more sophisticated" because it weighs poll results according to their rating of pollsters, but the ratings are subjective. Bottom line: helpful or not? Pick only one.]
But those who watch the averages closely may see what would seem to be two different situations right now. If you only followed RealClearPolitics averages, [HERE!] you would see President Trump’s job approval rating ticking up in recent weeks, from 41 percent approving as recently as July 18, to 44 percent approving as of Monday afternoon. In that same period, his disapproval rating dropped from 56 to 54 percent.
[I don't read 538 anymore because I got burned by them in 2016; because RCP was more accurate then; because one of 538's lowest-rated polls was Rasmussen, which was the single most accurate poll in 2016; because Nate Silver's goddamned probability percentages are hocus pocus that mean whatever Himself says they mean--Who knew that a 70% win probability meant that a race was essentially a toss-up, as Tarnished Silver explained in '16; because Silver is like Trump, or God, he takes no responsibility--"Why We Were Less Wrong Than the Others" was his post-Catastrophe mea culpa.]
Not surprisingly, that same RCP average showed that former Vice President Joe Biden’s lead over Trump has declined a bit. On July 18, Biden was ahead of Trump 49 to 41 percent. Biden ticked up to 51 percent in late July before returning to his current standing of 49 percent. Trump’s share, meanwhile, has ticked up to 42 percent, meaning the margin separating the two now stands at 7 points. [No, it was 6.4 on August 7, the date of Cook's article, and on August 5.]
If you looked across the Web at FiveThirtyEight, however, you would have seen an improvement for Trump—but a smaller one. On July 18, Biden led by 9 points, 50 to 41 percent; on Monday, Biden led by 8 points, 50 to 42 percent. Like the shift in the horse race that RealClearPolitics calculated, that marked a miniscule change...[I don't understand: it declined, by Cook's arithmetic, an identical 1% in both aggregates. A 1.6% decline would be 60% more serious, no?]
So, has there been any movement? If you look at polls in the current RCP average, covering July 21 to Aug. 2, you’ll notice that there is not a single poll that used live telephone interviewers. Rather, all are either online or interactive voice-response polls, aka robo polls. For the last few months, the online and robo polls have been a bit kinder to Trump than the gold-standard live-telephone-interview surveys. And of those online polls, none are from some of the superior entities that do that methodology, like the Pew Research Center.
[I did notice and caveated my evaluation accordingly.]
The last eight live-telephone surveys put Biden’s lead at 8 (a Fox poll conducted in July and a PBS/NPR/Marist poll), 10 (ABC/Washington Post), 11 (NBC News/Wall Street Journal), 12 (a Fox poll from June and Monmouth), 14 (New York Times/Siena), and 15 points (Quinnipiac). In terms of quality battleground-state polling, there again seems to be little movement.
...
The data suggest though that the tolerance that voters had for Trump’s unconventional style may have ended with his handling—or mishandling—of the pandemic, and the killing of George Floyd, and subsequent nationwide demonstrations.
[Point of order here: It has been comparatively very quiet these last few weeks. Trump's new campaign manager has managed to muzzle Trump. The most reputable polling firms using the "gold-standard" methodology tend to conduct their polls after something has just happened, right? When Trump was stepping in shit puddle after shit puddle the major's were polling after each step. A poll is a snapshot. If you only press the button when the guy is stepping in shit you are not taking snapshots of when he is not stepping in shit. It seems to me that is good reason to discount polls, even the highest quality ones, on temporality. I do discount them so. Which is why I wish one of them would poll during this relatively quiet period, but quality polling is expensive.]
It appears that among those outside his base, his credibility has taken a beating, his judgement and motives suspect. Trump’s ability to draw support beyond his base was predicated on keeping an economic tailwind that looks unlike to exist by November.
[Final, subjective, word is mine since this here is my own blog: Why no learned commentary on the state polls? Loaded question: because the state polls are even less reliable. We do not have a national election in the United States. We have fifty little elections. It is strictly a hypothetical point because it has never occurred in the history of the Great Republic that a candidate has won the national popular vote and lost in the Electoral College. RCP's aggregate of Joe Biden's lead in all twelve swing states *10:40 a.m.:just ticked up a tenth to 5.0. 2:40 p.m.: has gone up three-tenths of a percent to 5.2%. Two YouGov polls released today have Biden +6 in Pennsylvania and +6 in Wisconsin. Biden's advantage has always been less in the swing states that actually like decide the winner than nationally, which makes perfect sense. California, New York, and conversely Alabama and West Virginia do not swing. But even that one-tenth percent tick is likely polling artifact, right? The point is the swing state gap has gotten lower over the last little while. Fifty thousand or 70,000 or 80,000 strategically situated voters decided the 2016 election, with Russian assistance and without the "more sophisticated" Trump voter suppression initiatives that he has now. And Russia is better now. They have gone dark by moving their sites to the U.S. where the CIA is prohibited from monitoring. The undersigned has speculated that the swing state polls "undercount" Trump support by perhaps 200,000 due to these nefarious activities. Since 50k-80k resulted in the Catastrophe, since Biden is ahead of where HRC was in the swing states by 1.3%, might the Greater Rus-Trump Axis be able to erase Biden's lead in three or four of those states and result in Catastrophe Parte Deux? Methinks yes.]
Comes now the estimable Charlie Cook of the eponymous Cook Political Report who on August 7 advanced the polling axiom "garbage in, garbage out." In an article headlined, "The Polling Is Imperfect, but It’s Bad News for Trump" Mr. Cook writes:
...poll aggregates and models like RealClearPolitics and FiveThirtyEight.
...there are limitations to poll aggregations. Both of the main ones include surveys that I find awfully sketchy—and a couple that I find little short of fraudulent. Nate Silver’s FiveThirtyEight, which is far more sophisticated than the more raw RealClearPolitics, attempts to address the wide range in caliber of pollsters by assigning a rating for each pollster based on their track records. That has value, though I have real questions about many of the pollster ratings. I see some pretty mediocre firms with good ratings and some of the best in the business, many political firms whose polls are 99.9 percent of the time in the private domain, with lower ratings based on the very few polls that are publicly available.
[Mixed message: 538's aggregate is "far more sophisticated" because it weighs poll results according to their rating of pollsters, but the ratings are subjective. Bottom line: helpful or not? Pick only one.]
But those who watch the averages closely may see what would seem to be two different situations right now. If you only followed RealClearPolitics averages, [HERE!] you would see President Trump’s job approval rating ticking up in recent weeks, from 41 percent approving as recently as July 18, to 44 percent approving as of Monday afternoon. In that same period, his disapproval rating dropped from 56 to 54 percent.
[I don't read 538 anymore because I got burned by them in 2016; because RCP was more accurate then; because one of 538's lowest-rated polls was Rasmussen, which was the single most accurate poll in 2016; because Nate Silver's goddamned probability percentages are hocus pocus that mean whatever Himself says they mean--Who knew that a 70% win probability meant that a race was essentially a toss-up, as Tarnished Silver explained in '16; because Silver is like Trump, or God, he takes no responsibility--"Why We Were Less Wrong Than the Others" was his post-Catastrophe mea culpa.]
Not surprisingly, that same RCP average showed that former Vice President Joe Biden’s lead over Trump has declined a bit. On July 18, Biden was ahead of Trump 49 to 41 percent. Biden ticked up to 51 percent in late July before returning to his current standing of 49 percent. Trump’s share, meanwhile, has ticked up to 42 percent, meaning the margin separating the two now stands at 7 points. [No, it was 6.4 on August 7, the date of Cook's article, and on August 5.]
If you looked across the Web at FiveThirtyEight, however, you would have seen an improvement for Trump—but a smaller one. On July 18, Biden led by 9 points, 50 to 41 percent; on Monday, Biden led by 8 points, 50 to 42 percent. Like the shift in the horse race that RealClearPolitics calculated, that marked a miniscule change...[I don't understand: it declined, by Cook's arithmetic, an identical 1% in both aggregates. A 1.6% decline would be 60% more serious, no?]
So, has there been any movement? If you look at polls in the current RCP average, covering July 21 to Aug. 2, you’ll notice that there is not a single poll that used live telephone interviewers. Rather, all are either online or interactive voice-response polls, aka robo polls. For the last few months, the online and robo polls have been a bit kinder to Trump than the gold-standard live-telephone-interview surveys. And of those online polls, none are from some of the superior entities that do that methodology, like the Pew Research Center.
[I did notice and caveated my evaluation accordingly.]
The last eight live-telephone surveys put Biden’s lead at 8 (a Fox poll conducted in July and a PBS/NPR/Marist poll), 10 (ABC/Washington Post), 11 (NBC News/Wall Street Journal), 12 (a Fox poll from June and Monmouth), 14 (New York Times/Siena), and 15 points (Quinnipiac). In terms of quality battleground-state polling, there again seems to be little movement.
...
The data suggest though that the tolerance that voters had for Trump’s unconventional style may have ended with his handling—or mishandling—of the pandemic, and the killing of George Floyd, and subsequent nationwide demonstrations.
[Point of order here: It has been comparatively very quiet these last few weeks. Trump's new campaign manager has managed to muzzle Trump. The most reputable polling firms using the "gold-standard" methodology tend to conduct their polls after something has just happened, right? When Trump was stepping in shit puddle after shit puddle the major's were polling after each step. A poll is a snapshot. If you only press the button when the guy is stepping in shit you are not taking snapshots of when he is not stepping in shit. It seems to me that is good reason to discount polls, even the highest quality ones, on temporality. I do discount them so. Which is why I wish one of them would poll during this relatively quiet period, but quality polling is expensive.]
It appears that among those outside his base, his credibility has taken a beating, his judgement and motives suspect. Trump’s ability to draw support beyond his base was predicated on keeping an economic tailwind that looks unlike to exist by November.
[Final, subjective, word is mine since this here is my own blog: Why no learned commentary on the state polls? Loaded question: because the state polls are even less reliable. We do not have a national election in the United States. We have fifty little elections. It is strictly a hypothetical point because it has never occurred in the history of the Great Republic that a candidate has won the national popular vote and lost in the Electoral College. RCP's aggregate of Joe Biden's lead in all twelve swing states *10:40 a.m.: