Samples are no longer random or reprsentative and a variety of subjective and questionable methods and assumptions are used to make them "represntative."
Samples are no longer random or reprsentative and a variety of subjective and questionable methods and assumptions are used to make them "represntative."
It is a different kind of probability than "coin flip" probability in that each event is a unique event, so looking back at prior results is not that useful, if even useful at all.
Margins of error are too small and the stated confidence levels are too high.
The reality is that no one knows who is going to win with anything approaching accuracy and that the only thing you can say about the election is that it is too close to call on the available data. Anything else--probability numbers, etc is basically BS.
Klass outlines 4 possible outcomes (quote):
This assessment is based on an array of factors, but there are four major assumptions I’m relying on most.
First, I believe that pollsters aren’t accurately capturing the electorate post Dobbs (the Supreme Court decision that overturned Roe v. Wade and drastically reshaped the landscape of American politics). If the turnout model underlying the polling is even slightly off by gender and is undercounting low-propensity women voters who are motivated by abortion politics, that could prove to be a substantial source of unanticipated polling error.
Second, whether you agree with this strategy or not, Harris is making substantial attempts to win over traditional Republican voters who are put off by Trump. By contrast, Trump’s electoral strategy has largely been driven by an effort to fire up his base. He easily could have acknowledged mistakes or expressed contrition around January 6th, for example, but instead, he has opened rallies with the national anthem as sung by people who are in prison for attacking the US Capitol. He has also picked a VP, Vance, who acts like a laser-guided missile to alienate women. That’s not exactly a strategy to win back hesitant voters. Close elections are won by “base plus” strategies, in which the base turns out plus other undecided voters split for the candidate.
Third, having worked in campaign politics, I learned that the “ground game” matters. This refers to the field operations run by campaign staffers and volunteers, the human infrastructure that actually makes contact with individual voters, hoping to persuade them and remind them to vote. Every indication so far is that the Democratic machine is firing on all cylinders in the swing states.
Fourth, a new AP poll out yesterday places Harris at a +5 net favorable rating with registered voters, compared to a -18 favorable rating for Trump, a gulf of 23 points. In that same poll, Walz has a +3 favorable rating, compared to -15 for J.D. Vance. Yes, it’s one poll and polls are fallible, but that is a huge gap, unlikely to be explained away by statistical measurement error, and at some point, whether voters like the candidates or not does matter—at least a bit—in a close race.
I’ll put it a different way: there are basically four possible outcomes in this election:
Big Harris Victory
Narrow Harris Victory
Narrow Trump Victory
Big Trump Victory
My subjective belief, based on the available data and these four working assumptions, is that the first three outcomes are completely plausible. None of them would shock me. However, a Big Trump Victory—one in which he wins the popular vote by, say, five points and runs the table with the swing states would shock me. Meanwhile, I wouldn’t be remotely surprised if Harris won by five points.
If you want to understand polling better, you could a lot worse than reading this article (and the links to the actual research backing it up).
There is an EXCELLENT article on polling by Brian Klass over at the Garden of Forking Paths (Substack). It is behind a paywall though.
https://substack.com/home/post/p-150417111
TLDR version:
Samples are no longer random or reprsentative and a variety of subjective and questionable methods and assumptions are used to make them "represntative."
It is a different kind of probability than "coin flip" probability in that each event is a unique event, so looking back at prior results is not that useful, if even useful at all.
Margins of error are too small and the stated confidence levels are too high.
The reality is that no one knows who is going to win with anything approaching accuracy and that the only thing you can say about the election is that it is too close to call on the available data. Anything else--probability numbers, etc is basically BS.
Klass outlines 4 possible outcomes (quote):
This assessment is based on an array of factors, but there are four major assumptions I’m relying on most.
First, I believe that pollsters aren’t accurately capturing the electorate post Dobbs (the Supreme Court decision that overturned Roe v. Wade and drastically reshaped the landscape of American politics). If the turnout model underlying the polling is even slightly off by gender and is undercounting low-propensity women voters who are motivated by abortion politics, that could prove to be a substantial source of unanticipated polling error.
Second, whether you agree with this strategy or not, Harris is making substantial attempts to win over traditional Republican voters who are put off by Trump. By contrast, Trump’s electoral strategy has largely been driven by an effort to fire up his base. He easily could have acknowledged mistakes or expressed contrition around January 6th, for example, but instead, he has opened rallies with the national anthem as sung by people who are in prison for attacking the US Capitol. He has also picked a VP, Vance, who acts like a laser-guided missile to alienate women. That’s not exactly a strategy to win back hesitant voters. Close elections are won by “base plus” strategies, in which the base turns out plus other undecided voters split for the candidate.
Third, having worked in campaign politics, I learned that the “ground game” matters. This refers to the field operations run by campaign staffers and volunteers, the human infrastructure that actually makes contact with individual voters, hoping to persuade them and remind them to vote. Every indication so far is that the Democratic machine is firing on all cylinders in the swing states.
Fourth, a new AP poll out yesterday places Harris at a +5 net favorable rating with registered voters, compared to a -18 favorable rating for Trump, a gulf of 23 points. In that same poll, Walz has a +3 favorable rating, compared to -15 for J.D. Vance. Yes, it’s one poll and polls are fallible, but that is a huge gap, unlikely to be explained away by statistical measurement error, and at some point, whether voters like the candidates or not does matter—at least a bit—in a close race.
I’ll put it a different way: there are basically four possible outcomes in this election:
Big Harris Victory
Narrow Harris Victory
Narrow Trump Victory
Big Trump Victory
My subjective belief, based on the available data and these four working assumptions, is that the first three outcomes are completely plausible. None of them would shock me. However, a Big Trump Victory—one in which he wins the popular vote by, say, five points and runs the table with the swing states would shock me. Meanwhile, I wouldn’t be remotely surprised if Harris won by five points.
If you want to understand polling better, you could a lot worse than reading this article (and the links to the actual research backing it up).