It seems like both sides of this contentious election accept one fact as given, the poll averages can be trusted. I will explain why this is not true in 2016. Some pollsters are finding wild results that are inconsistent with other polling. Yes, pollsters will be judged based on their performance, but if many pollsters are making the same "mistake," then that mistake will be written off as understandable and not a consequence of bad polling methods.
Two main biases are hampering poll accuracy in my opinion, sample bias and the shy Trump voter. These two biases will combine to miss the surge of late support Trump will get and cause some polls to miss badly. Although a sizable number of pollsters will catch this surge, many of the pollsters who miss it will not care and in fact may wear it as a badge of honor.
A third bias, missing the new voters coming out to vote for Trump, is also real. Some traditional phone based pollsters will not count voters who did not vote in the last election as likely voters. This screen probably hurts Trump. It is harder to estimate this bias, however, so I will not be adjusting for it in my projections.
Evidence for Sample Bias
Most pollsters who use voter turnout models are assuming that the 2012 voter turnout will repeat in 2016. The Democrats outnumbered Republicans by six percent in 2012, 38D-32R-20I. Poll analysts are acknowledging that this could be wrong. In mid-September, Larry Sabato noted that this electorate may be leaning more Republican this time around, but he is sticking with the 2012 model nonetheless. That is the calculation most pollsters and analysts have stuck with.
It's also not a secret or a source of concern for them. Pollsters are using partisan distributions ranging anywhere from a +5 to a +12 Democratic sample on a regular basis. They believe that is going to accurately reflect voter turnout, perhaps because of the superior get out the vote (GOTV) operation of Democrats or because they assume it's the new norm for presidential year elections. Either way, they are boldly sticking to it, despite some strong evidence to the contrary. Young blacks and white millennials are not enthusiastic about this election and their turnout will likely be down. This would be sufficient to knock that +6 Democrat advantage to +3, which is the Democratic advantage I am assuming in my projections.
While Democrats will do all they can to get people out to vote, some desire on the voters' part is required to make GOTV really thrive. The reason GOTV worked so well for Obama among African-Americans and millennials was the historical nature of his candidacy. Black women represented the highest participating demographic in 2012. If that repeats, then maybe we will see a +6 Democratic turnout. I think that's quite unlikely. While Hillary would be the first woman president, it's not the same as the first African-American president and most people get that.
We are seeing evidence of seriously depressed turnout among blacks in early voting. While increased Latino votes can help Hillary in Florida, Nevada, New Mexico and Arizona, it's unlikely to make the difference because of increased white voter turnout. The one exception might be Nevada.
If the problem really is as pronounced as it seems to be in early voting in Florida, she could be in serious trouble in Pennsylvania and Michigan where there is not much of a Latino vote to offset it relatively speaking. Trump very well might need to win Colorado, Michigan or Pennsylvania to offset his loss of Nevada if Latinos save Clinton there.
The bottom line is that pollsters are ignoring polling evidence of superior Republican enthusiasm and depressed Democratic enthusiasm. I suspect many pollsters know that their turnout assumptions are unwarranted, but I also think they don't care. They have to make up for 2012 when pollsters really missed badly, and they did it in a way that was offensive to the politically correct elites. What better way to make amends than to dampen support for a candidate they consider anathema to the most rudimentary standards of politically correct decency?
This bias is reflected in all types of polls, including online polls. Because it relates to how the pollster arranges her or his data after collecting it, it doesn't much matter how it's collected.
In 2012 many pollsters underestimated black turnout and that requires penance in politically correct circles
Gallup predicted that Romney would win the election by one point, but Obama won by nearly four points. Gallup was so embarrassed and ashamed of its performance that it dropped out of polling the presidential horserace permanently. It wasn't just that they missed, they missed on the reelection of the first black president. Nothing illustrates the power of political correctness better than Gallup's self-imposed horserace polling death penalty.
Other respected pollsters that missed the election results badly include Mason-Dixon, who fared almost as badly as Gallup, and Monmouth, a pollster many consider among the best. The AP-GfK poll missed badly, as did Suffolk, the Field Poll and Rasmussen Reports. Other polls that did not do quite as badly, but still missed by nearly three percent over the last three weeks include the Washington Post/ABC poll, the Battleground poll, Fox News and NPR.
Collectively 2012 could be described as a polling train wreck. Pollsters lay blame on Hurricane Sandy, arguing that Obama received a late surge that went unnoticed. But the truth is that pollsters missed black and millennial turnout by a big margin and they are ashamed of that.
As this election has clearly demonstrated, the elite hold political correctness in high regard on both sides of the aisle. To fail to abide by the assumptions of the political elite represents a heresy that could cost you work, even if the cost is making turnout assumptions that are likely to cause you to miss the mark in your polling for one election cycle.
Doubt this? The big media is one of the biggest employers of pollsters and that media desperately wants a certain result. There is also a better than even chance that many of the big businesse CEOs that hire pollsters despise Trump and would be understanding of missing the 2016 election.
Simply put, pollsters have a built-in excuse for missing this election and they may even be rewarded for it by those who write the big checks. Pollsters can say they missed 2012 and needed to adjust and they just over-adjusted like most of their colleagues. Next, many of their potential employers, media and elite business, probably will discount their performance in this election because they agree with them that Trump needed to be stopped.
Evidence for the Shy Trump Voter
In addition to the evidence mentioned below, more recent polling evidence emerged for this phenomenon. Specifically, Republican pollster Trafalgar asked poll respondents in Nevada how their neighbors would vote after first asking how they would vote. The results gave Trump a 2.4% bump. While this particular survey included too many whites for the Nevada electorate, the survey nevertheless stands as evidence of the shy Trump voter. If Trump really does get an extra point or two out of this, it could easily put him over the top.
There is other polling evidence for this phenomenon, where a person who will vote for Trump won't admit it to a live pollster. Trump consistently does better in online scientific polls than in live caller polls. The theory is simple, that voter is embarrassed to admit a desire to vote Trump so he or she will not admit it. It's easy for them to say Gary Johnson or that they won't vote.
How many times have you wondered how so many people could remain undecided? Good question, because they aren't undecided. They just say they are. Many of them will for vote for Trump and they may even have a red hat on their dresser at home that they are keeping for a day when it might be acceptable to wear in public.
I imagine almost everyone reading this article has been tempted to deny voting for Trump or know someone who has. It is a remarkable phenomenon largely brought on by a hostile news media,
entertainment and academic culture. It is not only okay to bash Trump in those circles, it is cheered.
Nate Silver attempted to explain away the shy Trump voter by pointing to the GOP primaries. He noted that when comparing the primaries to the polls predicting the primaries, there was no evidence of a hidden Trump vote.
That is true, but it doesn't explain away the likely existence of the shy Trump voter. Voters who are going to vote for Trump but will not admit it probably didn't vote for Trump in the primaries.
They are people who supported Cruz, Jeb Bush and others who Trump had contentious battles with. The voters who supported those other candidates have two reasons not to admit they are voting for Trump, they are embarrassed and they don't like him. But despite those misgivings, many of these people are going to vote for him for the purpose of Supreme Court appointments. When you look at it this way, it becomes clear that there will likely be a hidden Trump vote and probably a big one.
This particular bias is confined to live call polls. Online polls will not see much, if any, of this bias.