The margin of error for polls six months out from election, if memory serves, is about 14%.
I think people are phrasing this wrong: it’s not that the polls are worthless, it’s that it does not tell you what’s going to happen on Election Day in any real sense. They’re useful for watching trends and gauging short term changes and impact. They are useful for telling you how things are going. They do not tell you anything remotely useful about how things will be.
Nor are they even remotely reliable to gauge things in the short term.
The methodology of collecting this data can be so heavily bias that the pollers can get whatever result they’re looking for, if they’re pursuing a narrative. I could write a poll that leads the poll takers to just about any desired conclusion by choosing very targeted questions with bad faith multiple choice options, and by conducting the polls targeting specific demographics. It’s a trivial thing to do.
Instead, you have to deep dive into the polling methodology, have a deep understanding of the quality of the poll operators, etc, to have any idea of if the poll was even trustworthy.
I, for one, dismiss polls entirely. There is too much disinformation, too many bad actors, whose entire goal is to “prove” their own biases in favor of their narrative, that the amount of shit buries the truth. So it seems a pointless exercise to sift through the shit to find the nuggets of truth, particularly when good faith polling isn’t at all reliable in the first place.
Exactly, also the expert in the article says basically the same thing in more diplomatic language:
However, speaking to Newsweek Todd Landman, a professor of political science at Nottingham University in the U.K., said it was “still too far out from the election” to read much into swing state polls.
He said: “The race remains highly volatile, and it is still too far out from the election to make any firm conclusion from changing polls across these swing states.”
Math is math. In order to calculate the margin of error you need to know the sample size. The number of months involved is not a part of the calculation.
Then it’s not margin of error, the predictive accuracy - whatever the term is - is far worse 6mo out from an election (5 now i guess) than the ones that are days or a week or so out. That’s the point. Polls now are useful but not for saying who will win in November. You may as well forget the top line numbers as soon as you see them unless you’re comparing them over time and/or looking at cross tabs for broad demographic trends, which is also limited but useful in some ways.
The margin of error for polls six months out from election, if memory serves, is about 14%.
I think people are phrasing this wrong: it’s not that the polls are worthless, it’s that it does not tell you what’s going to happen on Election Day in any real sense. They’re useful for watching trends and gauging short term changes and impact. They are useful for telling you how things are going. They do not tell you anything remotely useful about how things will be.
Nor are they even remotely reliable to gauge things in the short term.
The methodology of collecting this data can be so heavily bias that the pollers can get whatever result they’re looking for, if they’re pursuing a narrative. I could write a poll that leads the poll takers to just about any desired conclusion by choosing very targeted questions with bad faith multiple choice options, and by conducting the polls targeting specific demographics. It’s a trivial thing to do.
Instead, you have to deep dive into the polling methodology, have a deep understanding of the quality of the poll operators, etc, to have any idea of if the poll was even trustworthy.
I, for one, dismiss polls entirely. There is too much disinformation, too many bad actors, whose entire goal is to “prove” their own biases in favor of their narrative, that the amount of shit buries the truth. So it seems a pointless exercise to sift through the shit to find the nuggets of truth, particularly when good faith polling isn’t at all reliable in the first place.
Exactly, also the expert in the article says basically the same thing in more diplomatic language:
What horseshit… you need to know the number of people polled in order to know the margin of error.
I mean Larry sabato just cited this stat days ago but I’m sure you’ll say he knows nothing.
You can average the top performing polls to get this.
Math is math. In order to calculate the margin of error you need to know the sample size. The number of months involved is not a part of the calculation.
Then it’s not margin of error, the predictive accuracy - whatever the term is - is far worse 6mo out from an election (5 now i guess) than the ones that are days or a week or so out. That’s the point. Polls now are useful but not for saying who will win in November. You may as well forget the top line numbers as soon as you see them unless you’re comparing them over time and/or looking at cross tabs for broad demographic trends, which is also limited but useful in some ways.
Fair enough… if we both agree that “margin of error” has nothing to do with number of months; I have no argument.