"Shy" Trump vote stronger in 2020, only revealed by impersonal questions that let respondent avoid feeling shamed by media hysteria

Building on the previous post about how the polls are still wrong due to the warping effect of media hysteria, let's take a look at how the form of the question massively alters the outcomes -- from an apparent Biden landslide to a dead-even race.

There are really only two polls worth looking at this time, and they're the same two from last time -- the USC poll and the IBD poll. They were the most accurate, and they're tracking polls, which means they're sampling the same people over and over, not recruiting new samples every time. That eliminates response bias as a potential corrupter.

Response bias occurs when the people who do vs. do not agree to participate in the poll also differ in how they would answer the questions. For example, if Trump has a bad debate performance, it will demoralize his supporters and make them less inclined to answer the poll, while emboldening those of his rival and make them more inclined to answer.

If you recruit your sample after this influential event, it will bias the outcomes toward whoever the event cast in a better light. By recruiting a single sample in advance, and questioning that same group over and over, you're measuring true changes in their preferences, rather than people with some traits being weeded out and those with other traits being attracted to participate.

However, that still leaves the large problem of how forthcoming the participants are in their responses. Normally they wouldn't have to worry about telling the truth, but given the absolutely psychotic climate of hysteria that the media has created over the past 4-5 years, there's now an immense pressure to give the cultural elite approved answer -- i.e., against the fascist Nazi dictator Trump.

The media have made that portrayal so widespread that it's a background assumption of their poll questions. Namely, "Are you going to vote for the fascist Nazi dictator, or the guy who is not a fascist Nazi dictator -- just curious?"

Some people feel fine defying the fake news media, taunting them to their face at Trump rallies and so on. They will answer truthfully, denying the premise of the pollster's question. "He's not that -- that's just your BS propaganda about him -- and yes I am voting for him."

But others are not so comfortable with confrontation, especially if they're up against a representative of an elite sector like the media. They don't want to be mistaken for one of those uncouth types who screams "fake news," or they don't feel like getting dragged into yet another conversation about Trump that will not only go nowhere, but will result in them feeling shamed by the other side. These people are not going to open up, and will either give a non-committal answer like "don't know / third party," or will straight-up lie and say they're voting for Biden.

Given how hysterical and judgmental the cultural climate has become, I wouldn't be surprised if most of these types are answering "Biden" rather than "don't know / third party". If you give the non-committal answer, you're not really doing your fullest to stop the fascist Nazi dictator, and "don't know" implies you could actually vote for the Orange Bad Man by election day.

If you really want to be cleared of the charges brought against you -- "you're not one of those evil Trump voters, are you?" -- and not feel the spotlight of shame beaming down on you, you have to give an affirmative response for the anti-fascist Democrat superhero (the guy who voted for NAFTA and the Iraq War, and whose brain is melting out of his nose).

If the problem were just non-committal answers, then it would understate Trump's support but have no bearing on Biden's support. If the problem is saying "Biden" to escape the inquisitors unequivocally, then it both understates Trump and overstates Biden, warping the point gap between them even worse.

To overcome this obstacle created by the media, some pollsters, including the two good ones, have added questions that don't ask about the respondent's personal choice, which might activate that impulse to avoid or appease the inquisitors. Rather, they ask who the respondent expects to win among their social circle, or among their entire state.

This displaces responsibility for a potential Trump win away from the individual being questioned, and onto a diffuse group of others who cannot be identified for interrogation of their own. So if Trump wins, it wasn't me -- it was the Boogedy Voter! Don't punish me, go look over there in the shadows for the culprit, I know he must be lurking there somewhere!

Not surprisingly, these de-responsibilizing questions yield much more accurate answers. In the USC poll for Oct 11, their standard question (your choice, weighted by likelihood of turnout) produces Biden 54 vs. Trump 41 -- an utter wipe-out. How about your expectation for those in your social contacts? Suddenly the gap gets cut in half to Biden 52 vs. Trump 45. How about what you expect of those in your state? Now it comes down to a coin-flip with Biden 48 vs. Trump 46, statistically indistinguishable.

In fact, that question about those in your state has shown results that are statistically insignificant, because they're basically overlapping, for the entire tracking period back to Aug 17. Just like in 2016, this has always been a tight race for the pseudo-popular vote, and nobody is going to win it by big margins (over 5 points). It's likely Biden will wind up a bit ahead in the pseudo-popular vote, while losing the actual election to the first-term incumbent from the dominant party of the current period, beginning with Reagan.

(To see these charts for yourself, click on "All Graphs" near the top of the USC site, and mess around with the drop-down menus.)

The key thing for now is how drastically different the results are depending on the form of the question -- crucially, how tightly focused the spotlight of shame and blame is on the respondent. The standard questions produce risible outcomes -- predicting the unprecedented, while ruling out the amply attested. The de-responsibilizing questions produce outcomes that are within the realm of possibility.

That proves that it's a social pressure on the individual that's warping the picture in the standard questions. Anyone remotely in touch with the media of the Trump era knows what that pressure is, where it's coming from, and why some groups would want to escape its force.

We may have a priori guesses about which demographic groups would be most susceptible to these pressures from the elite media, but we can always look empirically and see which groups show the most risible results in the standard-question polls. For that, we'll look at the cross-tabs in the link above to the IBD / TIPP tracking poll, in a follow-up post.