NYT/Siena Polls of 6 Battleground States

[Dr. Arch]

Interesting how different narratives combine to spur multiple page discussions depending on the poll.

One thing that most posters here seem to be forgetting is that, on top of the uncertainty of the sample (MoE), there are also methodological assumptions being made on each poll's model of the electorate.

A model based on 2016 might not work in 2020, and those models may fail to capture unregistered voters or unexpected changes in the composition of the electorate on the day of the election (e.g. more voters from younger generations show up, Democrats that didn't vote in 2016 in the Upper Midwest show up, etc.).

TL;DR: Don't read into this as more than a snapshot of a possible result from this electorate model if the election were held today. Barring polls that show extreme certainty in the outcome (consistent large margins), we should save ourselves the time and wait until the day of the election, rather than wasting time like this bickering about polls a year out.

[Dr. Arch]

↑ Again, everyone is having a discussion about these polls without even knowing what the electorate will actually look like in 2020. It could look like 2016, or it could not. It's more likely that it won't because a lot changes in four years to keep that as steady as some may hope.

The electorate could very well be more educated (by function of lower non-educated turnout), younger (by function of younger voters not being captured by models, as they have tended to have lower turnout thus far), etc. and in numerous combinations thereof.

And this is all a year from the actual election.