Polling in America has always been used to assist the public’s and the candidates’ general feel of how certain elections may proceed. It can be used in a variety of ways, from indicating which states to focus your campaign in, to knowing which states you do not need to bother visiting, while also providing optimism for your side and/or pessimism for the other side. Since the ‘80s, the majority of Americans have always accepted the polls as accurate and useful tools to glance at for making predictions on who will win an election and where specifically those wins will be had. While all Americans knew that the polls were not infallible, we all gave the benefit of the doubt that they were trying to work in the best interest of the nation and provide accurate statistics for the rest of the country to follow.
This thinking now needs to be re-evaluated. Since 2012 and amplified in 2016, polling has become increasingly inaccurate to the point where margin of error no longer explains these huge mistakes in data. One could simply shrug these inaccuracies off as incompetence or a minor flaw in data, and that may be valid if this were a single polling firm or two. However, when this issue persists along every single polling outfit, something needs to be changed or investigated because this kind of misinformation cannot be propped up as “valid” if the trends of polling errors are getting progressively more egregious as the years progress. To underscore my point, let’s look at some of the polls themselves.
It is no surprise that the 2016 election was a huge upset against Hillary Clinton, for most polling firms and news agencies gave her an almost unbeatable lead against Donald Trump, with the New York Times giving her a 85% chance and the Princeton Election Consortium giving her a 99% chance of winning. To the pollsters’ credit, the national polling average for the 2016 election was pretty close, with the average being 3.6% and the actual being 2.1%. It’s when we get to the battleground states, however, where things start to get worse. Wisconsin was predicted to be a Hilary win by seven percentage points. Trump won it by 0.7%, well outside the margin of error. Iowa was predicted to be close, having Trump at +2.6%. He won by 10 percentage points. Michigan was predicted to be +6% for Clinton. Trump won by 0.3%. The list continues, but for brevity’s sake, we will leave it here. If these errors were within a couple of percentage points it would be understandable; however, having gaps of 6% or more is completely inexcusable and, frankly, disappointing.
It would be one thing if these polls were just a freak occurrence for that year in particular. After all, it was pretty well established that there was a “shy Trump vote” that the polls did not account for, so we should expect the polls to improve in 2020, right? Well, not only have the polls not gotten better, but they have gotten even worse. For example, back in 2016, the national polling was only off by a little over 1%. In 2020 (as of writing this with almost every state called), Joe Biden was predicted an 8% lead in national polling and only mustered a 3% lead, a five-point differential, which is insanely inaccurate for the most crucial of polls. In Florida, Biden was favored by 2.7%, and Trump won by 3.3%, a six-point differential. In Iowa, Trump was favored by only 1.6% and won by 8.2%. Trump was predicted to beat Biden by 0.9% in Ohio, but won by 8.2%. Wisconsin was supposed to be a Biden landslide at 9.2%, but instead, was a very narrow race, only netting Biden a 0.7% win. Even in easily predictable states like North Dakota, which gave Trump a 20% lead, he won by 33.4%, a 13.4-point differential. Tennessee is by far one of the worst: a 9% lead for Trump predicted and a 23.3% win for Trump instead, resulting in a 14.3-point differential. In fact, this trend carries on for a scary amount of states from coast to coast, suggesting that something is very, very wrong within the polling industry.
As one last piece of evidence, let’s take a quick look at the Senate and House polling in 2020 to see if polling on a smaller scale is any better. In Kentucky, Mitch McConnell was expected to win by around eight points, with many pundits claiming it could be a close race. McConnell blew out his opponent by 20.4%, a 12.4-point differential. In the House, Republicans were expected to lose seats to the Democrats. However, as of writing this, Republicans have actually gained five seats in some key areas that they were not predicted to win in, completely debunking the polls that claimed otherwise. The most egregious poll, however, out of anything mentioned here or previously, is most definitely the Maine Senate race poll: Susan Collins (incumbent Republican) versus Sara Gideon. Up to and including Election Day, Gideon was projected to win by eight points over her competitor in a supposedly landslide race. Collins not only won, but won by 8.8%, a roughly 17-point difference. This is so completely wrong that there is no explainable way outside of intentional skewing or severe incompetency that a race could be predicted so utterly incorrectly, and neither of those choices give us reasons to trust these garbage polls.
Polling in America needs to improve and pollsters need to be replaced and updated. When polling errors are all pointing in favor of one direction, there is something that is clearly wrong with the system. Whatever system polling firms are using now needs to change. Questions need to be briefer. Shy voters must be reached and accounted for. Unfortunately, until these polls can prove that they can be anywhere within the realm of accuracy, we have no other choice but to ignore them.
Christopher Barns is a sophomore majoring in physics with a minor in computer science. cb943290@wcupa.edu