Polls fail to capture Trump’s lead [OC]

14 comments

1. Data is from [fivethirtyeight](https://projects.fivethirtyeight.com/polls/president-general/2024/pennsylvania/)
   Download button is at the bottom of the page.

   Tools: Python to code everything and I used [Pandas](https://pandas.pydata.org/) to clean the data and [Seaborn](https://seaborn.pydata.org/)to make the graph.

2. I remember seeing an ad from KH near the end saying “don’t trust the polls, I’m getting washed. Please donate”

3. And this is with their models adjusting for unknown trump voters already.

4. All I can think of is how much the ones who got closer are going to upsell the shit out of themselves.

5. >The other possibility is that Trump surged late and that it wasn’t captured in the polls. However, this seems unlikely. And I can’t think of any evidence for that.

   [https://www.realclearpolling.com/polls/president/general/2024/trump-vs-harris](https://www.realclearpolling.com/polls/president/general/2024/trump-vs-harris)

   Clear upward trend for Trump from August to November with a drop for Harris in the last two weeks.

   Also, do your polling #s account for the fact that many polls have an option for undecided, but the election results do not?

6. I would be interested to see if they are consistently off across all three elections.

7. Because polls and surveys only work if participation is 100%. This is what people don’t understand. All these numbers are run based on full awareness. Never underestimate peoples ability to not do what they’re supposed to when inconvenienced. Polls mean nothing if it’s raining outside, or there’s traffic, or the chicken you were going to cook for dinner went bad two days ago.

8. I wouldn’t be surprised if something as simple as changing a screening question from “Are you *planning* to vote in the upcoming election?” to “*Are you* voting in the upcoming election?” would make the data more representative.

   Roughly — what? — 15 million Democrats just didn’t show up to the polls was it?

9. He *did* surge late though. It’s both pollster still not being equipped to deal with the inherent mistrust against them that anyone voting for Trump has, but there also was a surge that all pollsters took note of. You can, in fact, see it in the very link you posted, with Harris being almost 2% ahead of Trump in late September to suddenly being tied throughout all of October. That was a pattern across all swing states, all of a sudden it seemed that Trump was slightly ahead everywhere. This prompted worried democrats to appeal to potential “hidden” [female Harris voters](https://www.youtube.com/watch?v=Mm6Md87sE1U) that never materialized.

   That said, the big contributor is clearly pollsters underestimating support for Trump yet again.

10. “Silent” voters. People are either lying in polls are just simply not answering when their pick was ultimately Trump. I think it worked the other way too – except they may have been vocal Harris supporters and then just didn’t show up.

11. I have this suspicion that right-wingers are telling their voters to not share their opinion in polls, making it harder to gauge the mood.

12. For me, because of the threat of getting malware I never open and always delete no matter what party.

13. For me, and based on my experiences as a Turkish citizen, the pollsters all work for different lobbies and they care more about showing skewered results to discourage voting for other candidates than to reflect real results. Polling is already an extremely sensitive thing, you are polling a couple of thousand people to infer millions of people. Add to this the fact that the editor or the owner of the polling company wants a certain candidate to win, and thinks telling people that the other person is winning is bad, inefficient propaganda.

14. I’m convinced there’s a bias toward MAGA because the people polled on it are either ashamed and lie or lie for the sake of lying because they want to trick/surprise the libs.

Comments are closed.