- Joined
- Jun 20, 2008
- Messages
- 106,618
- Reaction score
- 98,442
- Gender
- Male
- Political Leaning
- Independent
My biggest problem with polls...and the first thing I look at...is their sampling data for political party. Many times one party or another is grossly oversampled. I consider that a useless poll and automatically reject that poll.
Regarding poll average sites. The problem is, if that polls that are included in the average mostly grossly oversample, the average is useless. We saw this during the last election and a lot of people were so surprised when the polls turned out to be wrong...they cried.
So...if a poll gets their sampling right, then I look at other factors such as what questions they are asking.
Sometimes...not very often at all...I come across a poll that is credible.
538 adjusts polls based on lean. Also, the problem with the polls in 2016 was twofold:
1. Polls record how people feel on a topic, but not what they will do. So it's easy enough to say that you hate trump and hate Clinton less so, but what the poll you're reading doesn't take into account is a depressed or energized result on election day. When I look to November this year, I'm looking at two things: the polls and the energy of Democratic voters, which is a hell of a lot different now than it was in 2016.
2. The problem with polls in 2016 was primarily about how people read them. So say a polling average says that 82% of people prefer Clinton over Trump. in statistics-speak, that wouldn't mean that Clinton will win, but that out of 100 elections, Clinton will win 82 of them. Well, that still means that out of 100 elections, Trump will win 18 of them. We happen to be one of the unfortunate 18 parallel earths.