The "shy Trump voter" hypothesis is not that there are lots of people who won't answer polls, but that there's a bias in the response rate and truthfulness of the responses received, and that pollsters are not able to accurately correct for that bias. So whether or not you answer polls doesn't speak to the validity or plausibility of that hypothesis. What is relevant is only the bit you say "might plausibly lean toward Trump"; clearly, your a priori opinion of that hypothesis is also fairly low.
> We've had four years of the media relentlessly telling us how stupid, evil, and wrong Trump is, and how stupid and racist all his supporters are. We've seen the media twist the facts to support this narrative.
You're talking specifically about the media that Trump supporters largely avoid watching and do not give any credence to. There's plenty of other evidence that Trump voters aren't ashamed of publicly expressing their support, and not much reason to believe that they would be much more shy about it when answering polls than when going about their daily lives.
The LA Times poll was one of few polls that predicted a Trump victory in 2016, the link has some details on the methodology:
https://www.latimes.com/politics/la-na-pol-trump-polls-20161...
Basically, if you ask someone what they think about police brutality you might get ‘of course it’s wrong’, but veiling that question a bit into ‘how important are safe neighborhoods, strong law enforcement?’ will take you closer to the truth.
Not everyone is happy with the riots in America at all, they view it as crime and certainly don’t want that stuff anywhere near their neighborhoods. Or further, they might not want certain types in their neighborhoods. You see where I’m going? If you start to probe in this way, you’ll find your hidden Trump voter versus asking ‘Did Trump handle coronavirus well?’, which everyone well answer no to, but if the person doesn’t want black people in their neighborhoods as a hidden belief, your polls are going to be fucked (since that’s a closet Trump voter).
No one will ever say ‘social welfare is unfair because minorities leech off of it’, only your most shameless republicans will say it. But guess what, people believe this deep down. How can pollsters surface that belief without arousing suspicion?
Way too many states in the rust belt went red to act like some negligible dynamic was in play in 2016.
Just to note, I am a liberal, in case anyone thinks I harbor those beliefs. But how would you know? :p
Now, do people like me lean one way? "Get out of my face and leave me alone" people might plausibly lean toward Trump, but I wouldn't be dogmatic about it. Even more, I wouldn't claim that it's statistically significant.
Second, you said:
> It is completely unreasonable to postulate that the effect would be drastically bigger this year, after four years of Trump actually being in power and normalizing his status.
Um, what? We've had four years of the media relentlessly telling us how stupid, evil, and wrong Trump is, and how stupid and racist all his supporters are. We've seen the media twist the facts to support this narrative. (Yes, Trump does so also. That's not the point.) I would be astonished if there aren't a bunch of Trump supporters who aren't going to tell pollsters what they think. (No, I can't prove it. Even if it's true, I can't quantify it. But I bet that they're out there, and that there are a lot of them.)