Analysis: Even in a close election, random chance means polls should be showing a broader range of results. That raises the question of whether we’re in for another polling surprise.
Another problem with polls is that the poll takers were off the mark in both of the last two elections, generally towards the Democratic side. So some of them compensate not by modifying their methodology, but by goosing the numbers by the same amount in the other direction this time around. They might say “Hey, we underestimated the guy by 2% in this state last time, so let’s give him a 2% mulligan”.
If you know polling is an inexact science, and you were wrong consistently in one direction twice in a row, it is better for your reputation if you are off in the other direction this time.
The problem with polls is they try to determine who is or is not a “likely voter”.
Look at the polling out of PA, check the left hand column here, 22 polls, 18 are “LV” or likely voter, as opposed to RV or registered voter.
https://projects.fivethirtyeight.com/polls/president-general/2024/pennsylvania/
In most cases, one of the criteria for “likely” is “did you vote in the last election?”
So, um, yeah, about that…
https://www.bernama.com/en/world/news.php?id=2358865
Those 100,000 new voters would have been discounted by pollsters as unlikely because they did not vote in 2020.
That doesn’t even take into account kids who turned 18 since the last election.
Another problem with polls is that the poll takers were off the mark in both of the last two elections, generally towards the Democratic side. So some of them compensate not by modifying their methodology, but by goosing the numbers by the same amount in the other direction this time around. They might say “Hey, we underestimated the guy by 2% in this state last time, so let’s give him a 2% mulligan”.
If you know polling is an inexact science, and you were wrong consistently in one direction twice in a row, it is better for your reputation if you are off in the other direction this time.