I definitely think shy clinton is far more common than shy trump - for a lot of the same reasons as shy Obama was always more than shy Romney. Demographics are destiny - and Trump sealed his own fate a long time ago - and a lot of voters marked "unlikely" are going to come out in droves to vote for Clinton, especially in the hispanic community. White men think their wives are going to vote for Trump, the reason why they believe there is a shy trump effect, but their wives are really going to vote for Clinton. And of course, there are people at the ballot box, facing their possible decision of voting for Trump, and ending up not going through with it because of WHO HE IS. These are right-leaning independents or even Republicans who are "for" trump but have been fighting with the idea of voting for him because, well, he's Trump.
As I’ve said many times in these threads, a polling error due to the likely voter models being wrong isn’t a “Shy” effect. It’s the pollsters getting their sample selection wrong. A “Shy X” effect is when the pollsters are wrong not because they messed up the sample, but because they’re being lied to, because poll respondents don’t want to reveal their true voting intentions.
And again, it’s also not about people “being shy” in public, or not telling their neighbors or even spouses who they’re voting for. It’s about them not telling pollsters. I don’t see any way to predict which candidate would be more likely to benefit from that kind of effect. But I also don’t think it’s going to be that significant. My hunch is that the more likely source of pollster error will be bad sample selection rather than deception on the part of those being polled.
Another thing to take into consideration: ground game operational disparity. Could be potentially exacerbated by Trump underperforming his numbers due to his toxicity, poisoning the well with Republican voters with rhetoric about vote rigging, and doubling down on marginal constituencies with almost no campaign turnout structure to capture them.
