As others have said you'd have to assume most of the mainstream pollers are being paid good money to try and influence public opinion. Which ironically could backfire on them as lazy millenials think Biden is a shoe-in. Just like 2016. While Trump voters who might obtain their news outside the mainstream, hear of Trump supporters being shot in the head at point blank range and the mainstream saying nothing.
Polling is a bit of a bedbug for me. The reason for the ... interesting results is two fold.
1: Systemic issues with polling. The constant oversampling of democrats is often pointed at, but its not as simple as just "There are more democrats in the sample". Any polling will weigh in the over sample and counter weight it to bring it into line with registered norms. The problem is this:
Say you have 100 Democrats and 50 republicans polled. The counter weight is simple here, halve the democrats individual value. Alright. Now, in this poll, 30% of democrats said they'd likely vote, and 50% of republicans did. So now you have 30 likely democrat voters, and 25 republicans, Alright, they provide their weight, so you halve the likely voters for democrats making it 15/25, right?
No. No they don't. They factor in the population in their weighting, but often forget/are too lazy too/are just idiots and don't change the result of the percentages.
So they get data that says, in effect, "60% of democrats will vote, and only 50% of republicans". Now, rarely is it THIS blatant, usually they do weight it slightly, but the weighting ITSELF is all off. Or they fudge the numbers a little and you get 51% democrats and 50% republicans. Its not just "They over sample" its "they over sample, and totally hash correcting the over sample"
Now, the 2: is easier to explain. The Shy Trump Vote is very real. The above issue is usually sorted by some secondary or even tertiary levels of statistical bullshit which is basically on giant kludge. These flaws have been there for decades now, but they were never made super apparent until... the shy Trump vote.
Lets ignore over sampling for now, lets say we have 100 each Democrats and Republicans. Now lets add in 40 Undecided Voters. So we have 100D-100R-40U. Normally, its assumed in political polling that undecided voters will break evenly for candidates. so you'd end up with 120D-120R. But, with the Shy Trump Voter, you have a lot of Undecideds being Secret Trump (ST) voters. So in actuality the math is saaaay 100D-100R-10ST-30u. Which, assuming the break even, means the final tally is 115D-135R,
All this doesn't even factor IN people who end up voting for the party they are not registered as (Polling universally assumes all registered democrats and all registered republicans vote for their party, based on Likelihood of voting percentages).
So, in summary: The percentages of Liklihood to Vote are off, usually by about 4-5 %, the Shy Trump Vote means that undecideds won't break down evenly, and Never-Trumpers and #WalkAway people totally fuck the base math to begin with.
variety.com
