Polls have long been considered the mathematical equivalent of “make it or break it” in the political world. Recently, polls have come under increasing criticism for their unreliability, bias, and inability to predict political outcomes. Most infamously, almost every poll in 2016 indicated that Donald Trump had practically no shot at winning the presidency, much less winning states like Pennsylvania, Michigan, and Florida.
How Polling Began
At its core, polling is all based on public opinion. The first known opinion poll ever taken was during the 1824 presidential election through tallies of voter preference in North Carolina. The poll results showed that Andrew Jackson would win the popular vote over John Quincy Adams, which was proven correct and drew interest to the idea of polling. Polls continued to occur, though they were restricted to smaller elections until 1916. In the 1916 presidential election, The Literary Digest magazine started the first nationwide poll by mailing out postcards and tallying responses, predicting the next five presidential races accurately. By this point in time, political polling had become famous for its increasingly accurate predictions.
Statistics-Based Polling Prevails
While The Literary Digest was increasingly popular for its polling accuracy, the 1936 election proved to be its first failure. The Literary Digest’s polling data was based on responses from their readers who were overwhelmingly Republican. At the same time, researcher George Gallup used a statistics-based survey to predict Roosevelt’s landslide victory that year, a massive contrast to The Literary Digest. Gallup’s new methods and accuracy would eventually come to set the standard for mathematic and science-based political polling until the present.
Modern Polling and its Troubles
Typically, polls nowadays are run by various outlets. Newspapers (The Hill, The Wall Street Journal, The Washington Post), magazines (USA Today), polling organizations, data analytic firms (Rasmussen, Gallup, YouGov), and universities (Quinnipiac, Harvard) make up the majority of pollsters in the United States. Sites like FiveThirtyEight and RealClearPolitics also exist, aggregating and averaging all national polls for most elections. These polling aggregators tend to be associated with the most reliability, though radical exceptions occurred in the 2016 presidential election.
FiveThirtyEight and other individual polls were criticized heavily in 2016 for their failure in predicting Trump’s sudden win. FiveThirtyEight in particular gave Clinton more than a 70% chance to win, and its prediction was supported by The New York Times, which gave Clinton an 85% chance to win. The failure of pollsters to predict Trump’s critical victories in Florida, Wisconsin, Michigan, and Pennsylvania severely damaged the reputation of polling. Few polls believed Trump had a serious chance at a win, and media pundits on Election Day treated Clinton’s victory as a given before Trump’s shocking victories.
Trump’s lackluster performance on polls and massive electoral win in 2016 was credited to the existence of “shy voters,” voters who answer surveys as undecided or anti-Trump out of fear of repercussions from their peers. Trump has called these supporters part of his “silent majority,” voters who do not voice their political beliefs out in public. The statistical blunders in 2016 raised questions about the reliability of polls when it comes to controversial candidates, political polarization, and overwhelming media bias.
Depending on who you talk to, polls are either useless or vastly important looks into the future. Since 2016, many polling outlets issued apologies for their inaccurate figures on Election Day in 2016. Outlets have also begun to publicize the corrections made to their polling methods to regain legitimacy with the public. Ultimately, polling means very little in elections whose results depend solely on swing states. Candidates can lead in national polls for months and still lose on Election Day depending on how the votes are allocated in individual states. State-based polling is now seen as the most reliable form of polling in predicting the winner of the presidency, and this is unlikely to change as long as the Electoral College continues to exist.