On election day, will Trump or Clinton over or underperform their polling?

[Blackacre]

I'm going for a Clinton overperformance roughly in line with what we saw in 2012, so she'll beat her polling by 3 points. I attribute this to shy clintons, but also to the Dems' turnout operation and republicans being unhappy with their nominee and staying home. If the latter happens a lot, she might beat her polls by 5 or 6 and take senate candidates with her, but to stay on the safe side I'll stick with 3. That would give her a 9-10 point lead in the NPV

[Erich Maria Remarque]

Anecdotally, I know far more shy Clinton voters than I do Trump voters.

How do you know who's a shy Clinton voter or a shy Trump voter?  Do people actually tell you "If a pollster called me, I wouldn't be honest about who I'm going to vote for"?

I really don't get this fixation on "Shy X" voters.  "Shy" in this context means people who will lie to pollsters, not people who aren't vocal about their voting choice with their neighbors.  Why does anyone think there's going to be a big "Shy" effect for either candidate, and what leads you to believe you can predict which candidate that'll be a bigger deal for?

Is there any reason to believe that if the polls end up being off, it's going to be because of a "Shy X" effect moreso than a pollster problem with sample selection?

You are right; there will not be any way to verify it. But we might be able to see some indication. If there is a Shy Trump effect, what type of voters will probably be affected most? Likely, a least hackish, but most educated and volatile type — college-educated and post-graduate Whites  [probably married]. So if we see some "strange compared to polls" Trump-friendly results in such an educated counties, it will be a pretty good indication.

You must be logged in to read this quote.

Because it did happen with similar to Trump candidates in Western countries, when their voters were described as "evil/stupid".

I was pretty sure about before, but I'm not any longer; Trump might cross the line lately, while Western populists had much more polished image and in general have had a good performance in the debates (both according to media and scientific polls).

[Mr. Morden]

You must be logged in to read this quote.

Because it did happen with similar to Trump candidates in Western countries, when their voters were described as "evil/stupid".

But again, do such polling errors occur because the pollsters reach a representative sample of voters, but some of those people lie about their voting intentions, or because supporters of certain candidates are less likely to pick up the phone and so the pollsters never even talk to them, or because the pollsters tend to screw up their likely voter models in races with such candidates, and thus aren't even contacting the right people to begin with?

A "Shy X" effect is at least the first of those options, though I think some people also intend it to mean the second even if they haven't thought about it in such terms.  The third though, about likely voter models, is not a "Shy X" effect as people are describing it.  But I'm not sure how you disentangle which of those effects is responsible for the polling error.

I'd also like to point out that the "Shy X" effect is always trotted out when there's a divergence between live phone interview polls and robo-polls or internet polls.  The implication is that people will lie to a human, but not to a recording, and therefore the live interview polls are less reliable.  Except...all indications over the past decade+ are that live phone polls are if anything more reliable.

[Yank2133]

I don't know why people assume Trump has more "shy" supporters. During the primaries, Trump had the loud majority and Clinton had the silent majority.

Not only that, Trump under performed his polling numbers in the primaries.

[Tender Branson]

Impossible to tell ...

In the Austrian presidential race, the populist far-right candidate vastly outperformed the polls in the first round, but the polls were pretty accurate in the runoff.