President Donald Trump has fallen recently in national polls, and there is great concern among some Republicans that he could lose his bid for re-election. Is Joe Biden set for a victory, or are the polls wrong?
In 2016, nearly all national polls gave Hillary Clinton a substantial lead over Trump in the weeks and days leading up to the election. A survey by Princeton Election Consortium three days before the election gave Clinton a 99% chance of winning and predicted 312 electoral votes. She ultimately lost and ended up with only 227.
The pollster who was most accurate in 2016 turned out to be Robert Cahaly of the Trafalgar Group. He called Trump’s victory, and he managed this feat by recognizing that Trump was getting support from unlikely voters.
Furthermore, Cahaly understood that many of Trump’s voters refused to provide an honest answer on the polls in fear of their jobs and lives. He added questions to his surveys that would reveal this group of voters.
Polls have been notoriously unreliable predictors of elections, especially when they are performed before the official campaigning starts. In July of 1988, for example, Michael Dukakis had a solid 17% lead over then-candidate George H. W. Bush in a Newsweek/Gallup poll. As history records, Bush came out ahead.
Only a few weeks before the 1980 election, the incumbent President Jimmy Carter was solidly ahead of then-candidate Ronald Reagan. Due to his performance in the debates, Reagan closed the gap and experienced a landslide victory.
A factor that is far more predictive than polls is voter enthusiasm. Trump enjoys massive approval within his party, and – even though his candidacy is uncontested – he got more votes in many of the primaries than Biden.
By contrast, a national survey by SSRS for CNN showed that 60% of Biden-supporters are voting against Trump, rather than for Biden.
Fast-forward to the present day, and the national polls do not seem to have updated their models to capture unlikely and silent voters appropriately. Cahaly says that the fear of voicing support for Trump, even in a supposedly anonymous poll, is even greater today.
Repeating his method from 2016, he finds that the race between Biden and Trump is much closer than the aggregate of polls implies.
The Primary model, which has accurately predicted 25 of 27 elections since 1912, gives Trump a 91% chance of winning in 2020, according to Dr. Helmut Norpoth.
The year of COVID-19 has proven how unreliable both experts and the media’s reporting of facts and statistics are. Election polls are likely no exception.
Trump may be down in the polls, but – based on the models that have proven accurate in the past – he is still well-positioned for re-election in November.