AlleyPack said:
Based on some quick Googling (and I'm just going to assume the numbers are fairly accurate):
-- Approximately 139 million people voted in 2016.
-- Over 93 million people have cast early votes so far in 2020.
So given the above info, exactly what are most of the polls basing their data on?
Actually speaking to people who have cast their early votes, or rather just random phone polls?
And then a follow-up question to those two above sets of numbers:
-- How many MORE people are expected to vote on Tuesday?
-- How many mail-in ballots will there be, nationwide?
(sub question) -- If some mail-in ballots have already been received by their respective states, have they counted towards that 93 mil "early voter" number?
I just don't see how ANY of these polls can have a truly firm grasp on what's going on, either way.
It seems way more "wishful thinking" and "intuition" than anything that's based on a full picture of the data.
Professional pollsters are just relying on wishful thinking and intuition? Their work would be borderline useless if that were the case. These unknowns that you're asking about are calculable.
The highest-quality polls historically are live caller phone polls. Pollsters are asking people who they're going to vote for and how likely they are to vote (or if they've already voted, who they've already voted for).
Voters that they reach that have not voted are assigned a probability of voting based on how sure they say they are that they'll vote. Voters that have already voted are normally assigned a probability of 1 to account for their vote already having been cast.
And regarding how firm their grasp is, it depends on what you mean by "truly firm." Their results are not ironclad, that's why there's always polling error.
Historically they're accurate enough that for presidential elections they're within +/- ~3%. In 2016, that was true nationally but at the state level several key states had polling error above 5% which played a key role in picking the likely winner. Trump went on to win several states he was a very small underdog in by less than 1% of the popular vote in those states. Basically that 5% error in several states swung the election his way.
Most of that error was attributable to very few state pollsters adjusting for education level and Trump being disproportionately popular with non-college-educated white men.
Supposedly, many/most state polls have tried to correct for that source of inaccuracy this time around.
We'll see if they did, or if some other source of polling error creeps in this go-round. This is the constant challenge for pollsters; if they could detect and correct for all potential sources of error in advance, there wouldn't be polling error.