The polls appear to be shifting in favor of Joe Biden.  A commonly heard explanation from Trump supporters is many of the polls are oversampling Democrats.  This is true in some cases but not all.  So, is something else going on here we do not really understand?

It is difficult to get an accurate sample of the electorate, especially when party affiliation numbers vary with the political climate.  According to a Gallup survey during the first two weeks of September, party affiliation comes in at Republican 29%, Democrat 30%, and Independent 40%.  When you factor in voters who lean one way or the other, the split is a 5% Democrat advantage, 50-45.  

Most of the polls in the Real Clear Politics National Average, at least the ones without a subscription in which I can find data on party affiliation, are fairly close to the Democrat +5 number.

• The Harris poll of registered voters gives the Democrats a 5% polling advantage, 37-32, and has Biden up by 5 points.
• Fox News gives Democrats a 7-point advantage 49-42, and has Biden up by 10.
• Reuters gives Democrats a 4-point advantage, 45-41, and has Biden up by 12 points.  The party split sounds reasonable, but these are registered voters, and 9% of their sample is not Republican, Democrat, or Independent, making them an unknown factor. 
• CNN gives the Democrats a 5-point advantage, 33-28, and has Biden up by 16 points.
• The latest Economist/YouGov poll favors Democrats by 10%, 37-27.  They have Biden up by 9. Oversampling of Democrats appears to be the game here.
• The New York Times poll is a surprise. It gives Republicans a 1-point advantage, 32-33, and has Biden up by 8.

The surface reality does not seem to reflect the pro-Biden polls.  If you believe the liberal press, everybody hates Trump, but he routinely draws enthusiastic rally crowds in the thousands. Multiple pro-Trump boat parades have drawn hundreds and thousands of watercraft.  Ditto for vehicle parades.

Everybody supposedly loves Biden, but his crowds are usually less than a hundred, and sometimes nothing more than the press corps.

There were at least Five million firearms purchased by first-time gun owners thus far in 2020.  I am sure they will all vote for Biden so he can take their guns away.

Biden has the support of somewhere around 175 law enforcement officers.  Trump’s support from law enforcement is estimated at 900,000.

According to Bloomberg, tens of thousands of small businesses are going bankrupt thanks to COVID-19 restrictions.  Hundreds more have been looted and burned in riots.  No doubt, all the business owners will support the Democrats who allowed this to happen.

According to a September Gallup poll, 56% of Americans believe they are better off today than they were four years ago.  They must all hate Trump, too.

Okay, how do we correlate the political polls with the facts on the ground?  More importantly, are the polls missing a hidden Trump vote?  And if it exists, can we quantify it?  To get an idea what is happening, we have to start with the present social climate.  

The level of discord and division in the country has only been matched or exceeded twice before.  Once during the Revolution and a second time during the Civil War.  Much of this has been generated by the news media and the Democrats, who have waged a four-year-long negative information campaign against President Trump and his supporters.  The social pressure created by voices screaming "liar," "racist," "white supremacist," etc. at a select segment of the population is having an effect.  This may be causing voters to either lie in response to some of the poll questions or avoid political polls all together.

According to a recent Cato Institute poll, 77% of conservatives are now afraid to express their political views versus roughly 50% of liberals.  This includes 40% of Republicans with a college degree and 60% with a post-graduate degree, who are afraid their political views will harm their careers, versus only 25% of Democrats.  If you felt your job was threatened by expressing your political views, would you lie in response to a poll question?

There is also a psychological phenomenon known as “social anxiety” that can be caused by the fear of being judged by others.  Social anxiety can produce something called social desirability bias, which causes people answering survey questions to respond in a manner they believe will be seen favorably by anyone who hears or reads the answers.  In other words, if you felt an honest response would make you seem to be insensitive, uncaring, or even racist, would you lie in order to be seen more favorably by the pollster?  If yes, you would be exhibiting social desirability bias.

This may explain why a Cloud Research poll taken between August 19-27 discovered roughly 10% of Republicans would not tell the truth to a political poll.  Although Cloud Research says the results cannot be applied to the voting population at large because of the poll’s design, they ran two different panels of 1000 people each, and got the same result.

So, we have valid reasons for Republicans to be untruthful and a poll suggesting a possible number.  Is there any evidence this is happening or has happened?  Actually, there is.

If you look at the Gallup party affiliation poll mentioned above, which includes independent leaners, Democrats and Republicans were tied in April 2020 at 46-46.  Only one month later in May, after the George Floyd incident, party affiliation widened to a 12-point Democrat advantage, 50-38.  Voters claiming to be Republicans declined by 8 points.  Where did they go?  A simple guess is due to social pressure 4% were claiming to be Democrats and another 4% were claiming to be Independents.  Since Republican party affiliation has now returned to 45%, the best guess is around 7% of Republicans were being untruthful in May.  One could argue events following the Floyd incident, including the riots, brought voters back to the party, but it is more likely they just needed an excuse to claim to be Republicans again.

What about now?  Is there a number we can point to and say “this is the silent Trump vote?”  No.  Picking a number would be nothing more than a guess.  However, we can judge the effect a silent Trump vote would have on the election.

In the following composite estimate, I am using the Real Clear Politics averages of each individual battleground state and competitive senate race, with one difference.  I am removing polls that are outliers and recalculating the RCP average.  Example: in Florida, the RCP average gave Biden a lead of +3.7 points.  I removed the Quinnipiac poll at Biden +11 and recalculated the average using all the other polls, which gave a more reasonable average of Biden +2.75.  Here are the results.

What you see in the table is the average level of silent support President Trump needs in the battleground states to win a second term, if the election were held today.  If the silent Trump vote is 2%, he has a chance of winning. At 4% it is likely Trump wins.  Between 2% and 4% it becomes a question of turnout.  The table also shows what Republicans need in the Senate races involving Collins, Tillis, Ernst, Loeffler, and possibly Peters in Michigan, to maintain control of the Senate. If the silent vote is 3% or higher, it is likely Republicans keep the Senate.  If polls tighten, the number of silent votes needed for victory will also go down.

Skeptics will tell you it is impossible for modern polling techniques to miscount this much of the electorate.  Pollsters claim their final prediction of the vote in 2016 was within the margin of error.  Yes, but this is not 2016.  I hesitate to use this comparison, but I believe I am correct is saying conservatives have endured a level of hate over the past four years similar to what the Jews endured in Germany in the 1930s.  But we do not have brown-shirted gangs roaming the streets, assaulting citizens, breaking windows, and setting businesses on fire… or do we?  How much social pressure has to be brought to bear before people fear to tell the truth?  We will find out in a few weeks.

Image: R. Nial Bradshaw