Another Polling Miss: How Pollsters Got it Wrong in 2016 and 2020

As this unusual election cycle comes to a close and pundits across the political spectrum crunch and spin the data that comes in, one area that always attracts appropriate scrutiny is the state and national polling that is relied upon by campaigns, the media, and the American people. Marred by controversy in the previous election cycle for getting the seemingly inevitable Hillary Clinton win wrong, the polling industry has taken steps to identify and acknowledge in a (rightly) humble way its past shortcomings as a quickly evolving electorate and shifting media consumption landscape keep pollsters playing catch up cycle after cycle. So, what was discovered after 2016, what factors were changed or unchanged, and what happened in 2020?

Much like 2016, pollsters got a lot wrong in 2020. Based primarily on opinion polls, election pundits put Clinton’s chance of winning somewhere between 70-99%, making her the reported frontrunner even in states that Trump would eventually carry like Wisconsin and Pennsylvania. In fact, all 30 polls taken in Wisconsin in 2016 had Clinton winning by anywhere from a 2%-16% margin of victory, according to Real Clear Politics.

One factor leading to the intangible that can be quantified as constant as there are bad polls every cycle are the ‘Undecided Voter’, those on the fence, last minute electors who appear like a deer in headlights on Frank Luntz focus group segments on Fox News. However, where pollster blundered was the small sample size pollsters attributed to this class of voter, which ended up being much higher than quantified and breaking for Trump.

There are other significant misses from 2016 that have also been acknowledged, such as under sampling non-college educated white voters, who broke heavily for Trump and turned out to vote in 2016 when they hadn’t in years past, hence the phrase ‘silent majority’ was born. Together, these factors gave election forecasters, pundits, and media personalities reason to say with false confidence their prediction victor.

In 2020, differences emerged leading to a slightly improved, but still missing mark, result in polls on the national level. State level polling, particularly in Rust Belt states, still left things much to be desired. For instance, 2020 Wisconsin pollsters for ABC News/Washington Post found Biden up 57% days before Election Day. Biden would underperform that metric by about 8%, more than double the polls stated margin of error.

While objective observers look for traditional indicators in poll results, such as a candidate’s job performance, the state of the economy, or the answer to the classic ‘Are you better off than you were four years ago’ question, there is more emotion to voting than pollsters would like to admit (their profession be doomed), even beyond the somewhat quantifiable “enthusiasm” barometer. This enthusiasm for Trump can manifest itself in multiple ways, including what pollsters are now pointing too in defense of their miscalculations, the ‘shy Trump voter’ who may just have purposely answered pollsters in such a way as to skew the results away from their truly preferred candidate.

These voters were “systematically under-sampled and undercounted,” according to Todd Eberly of St. Mary’s College. ‘“I don’t know necessarily what [pollsters] do to adjust for that going forward.”’

If pollsters are going to save their credibility in 2024, they need to spend the next four re-evaluating their assumptions about American voters and finding new ways to adequately measure public opinion. The mistakes of 2016 and 2020 are fixable, but only by those willing to admit they were wrong and need to start over.