Member-only story
2020 Polling Is Pretty Much Useless at This Stage (Except for TV Ratings)
Tune out the pundits. Polls showing Trump in trouble mean almost nothing this far from Election Day.
Do the polls really show Donald Trump is headed for big trouble in his reelection campaign? That’s what you might think from watching coverage of early polls of the 2020 general election. In reality, though, it’s just too early to learn much from surveys testing how Trump would fare against various Democrats. Unfortunately, media reports frequently fail to convey this uncertainty, which may lead people to underestimate Trump’s chances.
How do you fill air time more than a year before the general election? One common approach is to engage in breathless hype around the so-called trial heat polls that pit Trump against potential Democratic opponents. Right now, they typically show Trump losing. For instance, a Quinnipiac University poll released last week found Trump trailing behind numerous Democrats, including Elizabeth Warren (by 7 points) and Joe Biden (by 13 points). These data were widely covered on cable news: MSNBC’s Rachel Maddow described them as “brutal for Donald Trump” and CNN’s Rosemary Church talked at length about the “commanding lead” Biden seems to hold.
Yet studies show that national polls conducted 300 days in advance have virtually no predictive power. We’re still more than 500 days away.
Courtesy of the Roper Center for Public Opinion Research, here are just a few examples showing how wrong trial heat polls conducted in June of the year prior to an election can be:
- In June 1983, Walter Mondale led Ronald Reagan 49% to 39%.
- In 1991, George H.W. Bush was supported by 51% of Americans, compared to 28% who supported a Democratic candidate and 21% who said they didn’t know.
- In 1995, Bob Dole led Bill Clinton 48% to 44%.
- In June 2011, any Republican candidate was preferred to Barack Obama 44% to 39%.
We saw that same pattern in June 2015, when a CNN/ORC poll found Hillary Clinton leading Donald Trump 59% to 34% — not exactly a useful forecast of how the election would turn out.
