It’s November 8, 2016, and celebrations are already underway at Hillary Clinton’s election night party in Manhattan. Earlier in the evening, the New York Times reported Clinton had an 85 percent chance of winning. Supporters gathered under the glass ceiling of the Javits Center, waiting for their candidate to shatter the symbolic Glass Ceiling for women in politics. However, the only thing that would be shattered that night were their hopes.

Nearly three years after the 2016 U.S. Presidential election, the world is still reeling from what will perhaps go down as the most shocking political upset in American history, if not the history of democracy. On November 8, 2016, billionaire businessman and political outsider Donald Trump became President-elect of the United States, defying nearly every poll, pundit, and political prediction.

Well-known pundits, both Democrat and Republican, indicated Hillary Clinton was a clear favorite, a shoe-in for the presidency, even up to until a few hours before the results. In October 2016, poll expert Sam Wang even tweeted that he would eat a bug if Trump got over 240 electoral votes. The consensus in the media was so strong, a Trump victory seemed laughable. Moreover, developments in the months leading up to the election– including allegations of sexual assault, entanglements with a porn star, and the release of a 2005 tape in which Trump brags about grabbing women by their private parts– signaled his campaign was doomed.

Nevertheless, as election night wore on, blue states on the electoral map were slowly but surely replaced by a wave of red. Defying the polls, Trump won key states in which he had never been ahead, securing the 270 electoral votes needed to win the presidency. Overall, Trump got 306 electoral votes (and poll expert Sam Wang ate a bug on national television).

The mood at the Javits Center turned so grave, it would have made a state funeral look like a state fair. Clinton supporters appeared shell-shocked. People sobbed on television, college classes were cancelled, and a pending apocalypse was predicted, if not all but expected. Amidst the shock and sheer surprise of the election, many wondered how the polls could have been so wrong.

So, why were the polls so misleading? On election night, the national polls put Clinton at a 52% to 53% chance of winning the two-party vote. She ended up winning the popular vote by a margin of about 2 percent, meaning the error in the national polls was not significant. However, Trump’s unexpected success in flipping the key battleground states of Wisconsin, Michigan, Pennsylvania, and Florida– all won by Obama in 2012— gave him the majority of Electoral College votes needed to win the presidency.

A quick refresher: The Electoral College, the system by which the U.S. President is elected, was devised by America’s Founding Fathers to serve as a check on direct democracy and increase the power of small states, encouraging candidates to appeal to voters outside of large cities. The candidate who wins the majority of votes in a state generally wins all of the states’ electoral votes (winner-takes-all), even if he or she only won by a slim margin.

The main predictive failure in the 2016 Presidential Election was that most polling at the state-level consistently underestimated Trump’s support in the key battleground states. 206 counties, many located in the Rust Belt, unexpectedly flipped from voting for Obama in 2012 to Trump in 2016.

“National polls were pretty good in the 2016 presidential election, but state-level polling was fairly poor,” writes prominent statistician Nate Silver on his site, FiveThirtyEight. The fact that the national polls were “pretty good” is significantly secondary to the fact that state-level polls were “fairly poor,” since the U.S. President is not elected based on the national popular vote. So, what explains pollsters’ underestimation of Trump’s support in these key states?

One hypothesis that has been put forth to explain the discrepancy is the silent Trump vote, or the “shy Trump supporter” phenomenon, which polls would not have reflected. According to this hypothesis, social desirability bias (in this case, hesitance to express favor for an unpopular candidate) caused some Trump supporters to conceal their vote preference from pollsters. According to some scholars and pollsters, however, there is data to suggest that the “shy Trump voter” phenomenon was not in play.

“Trump outperformed the polls the most in states such as North Dakota and West Virginia where we assume respondents would’ve had little embarrassment in declaring their support for him, while he did no better than the polls’ predictions in solidly Democratic states,” write Andrew Gelman and Julia Azari in a paper published by Columbia University. Nate Silver presents the same argument in his article, “‘Shy Voters Probably Aren’t Why the Polls Missed Trump.”

However, these observations are not conclusive, as they don’t consider other factors, such as the discouraged-voter-effect, that could have come into play in solidly blue or red states, where political minorities tend to believe their vote won’t count much.

Even if voters did not intentionally mislead pollsters, which is very plausible, there is another development that substantiates the “shy voter” hypothesis. Two weeks before the election, the five key swing states that were predicted wins for Clinton– Pennsylvania, Florida, Michigan, Wisconsin, and North Carolina– recorded a historically high percentage of voters who said they were either undecided or would vote for a 3rd-Party candidate.

According to FiveThirtyEight’s national polling average, the number of voters that polled as undecided or voting for a third-party candidate was much higher in 2016 than in recent elections:  In 2016, 12.5% of voters polled as Undecided/Other, compared to 3.1%, 3.4%, and 3.7% in 2012, 2008, and 2004, respectively.

“Third-party voters often have a weak commitment to their candidates and can be picked off by one of the major party candidates,” writes Silver, who assumed Clinton would win most of those Undecided/Other votes. Instead, they mostly went to Trump in all five states and pushed him over the edge in the Electoral College vote.

Can we assume that the increase in “Undecided/Others” voters was a result of the “shy Trump voter” phenomenon? There is no way to know for sure, but it is certainly plausible. It’s also possible that the anti-Trump sentiment fueled by the media contributed to low response rates, or survey-shyness, among Trump supporters. This would also explain why Trump’s supporters were under-polled in virtually every poll in the five key states.

“When there is good news about a candidate, his or her supporters are more likely to respond to polls,” write Gelman and Azari. “The final polls were off by about 2 percentage points, suggesting that, even at the end, Trump supporters were responding at a lower rate than Clinton supporters, most notably in certain key swing states.” Moreover, the numbers didn’t need to be significant to significantly skew the polls and change the election.

Now, let’s look at some previsions for 2020. According to state-by-state Gallup polls released in February 2019, Trump’s approval ratings in the key states of Pennsylvania, Wisconsin, and Michigan– which largely accounted for his victory in 2016– are not strong. 42% in these states approve of the job Trump is doing, and in all three, a majority disapprove. Already, a new poll has come out indicating that Trump is trailing behind Democratic candidate Joe Biden in the key battleground state of Pennsylvania.

However, if some form of the non-response bias was at play in 2016, we can assume it will be at play in 2020. In swing states, silent supporters matter. They only need to exist in small numbers to make a significant difference in the polls and at the ballot box, as we so shockingly learned.

There is also a second, perhaps more prevalent, kind of silent supporter. This isn’t just the supporter who is silent in the polls, but the supporter who is silent in daily life and in the public dialogue. A May 2019 Gallup poll reveals that Trump’s approval rating has hit 46%, the highest since his first month in office. Though that is relatively low when compared with the averages of other presidents, it’s high for Trump, and it is 2 percent higher than Obama’s was at the same point in his presidency. Meanwhile, the mainstream media is still rife with disapproval, and in cities like Los Angeles, New York, and San Francisco, expressing support for any of Trump’s policies is taboo, if not entirely socially prohibited.

A study conducted by Harvard University’s Shorenstein Center reports that the tone of Trump’s news coverage during his first 100 days in office was 80 percent negative and 20 percent positive. According to Pew Research, news stories about Trump were only 5 percent positive in 2017, compared with 42 percent positive for Obama in 2009.

“In the press and in the academy, Trump is almost uniformly recognized as a catastrophe, the worst president in history,” writes David Graham in The Atlantic. “Trump’s flaws have been so extensively cataloged that it’s easy to lose sight of his strengths as a politician… [He] possesses an unequaled instinct for connecting with voters and exploiting their grievances.”

Many of these grievances are economic in nature. Civic life is not as flourishing as it was in the days when Tocqueville wrote of democracy in America. When it comes to politics, people care mostly about their own interests. Economic well-being is usually top of mind. And Trump certainly has a booming economy on his side, with strong job growth across the board and an increase in US GDP. A recent CNN poll reported that 71 percent of Americans view the U.S. economy as “very good” or “somewhat good,” higher than has been measured since 2001.

“Despite the fact that Trump is a largely incompetent clown, Democrats should not be overly confident or sanguine that they can beat him,” Dan Pfeiffer, a top aide to former President Barack Obama, told Politico. “He is a slight favorite to win. But he barely won last time and it took a Black Swan series of events to make that happen.”

The idea that it was merely a Black Swan series of events that led to Trump’s victory is a dangerous one. We need to remember that even though Trump was opposed as dangerous, uncouth, and overall unfit for the presidency by almost everyone in the establishment– including most of his Republican primary opponents at some point, many conservative intellectuals, and nearly the entire media– almost half of voters chose to go with him anyway.

While Trump’s victory has been largely dismissed as a fluke, the result of an unpopular democratic candidate, a racist outlash, and even Russian interference, some of the issues he tapped into deserve consideration. Trump sent Americans a message that resonated with them, a message that his Republican counterparts could not deliver for fear of being ostracized. The three main issues that catapulted him to the presidency– opposition to the establishment elite, trade, and illegal immigration– are not going away any time soon. Moreover, Trump has been dogged in pursuing his key campaign promises, even when they seem to undermine his popularity with establishment Republicans and the nation at large.

To allow for a flourishing, civic society we need to be able to have balanced conversations about important issues that affect Americans on all sides of the political, economic, and social spectrum. Criticism of our president is necessary, and often warranted. The problem is not that there is criticism of Trump or his character. The problem is that some of the legitimate issues he seized on are contemporaneously thrown under the shade of the “terrible Trump” umbrella, so that honest conversation about them is virtually impossible. This blocks civic dialogue, instills hatred on both sides, and causes the silent supporters to grow even more silent.

Donald Trump might not win easily in 2020. In fact, he might not win at all. But if he does, it is likely that the silent supporters– those not reflected in the polls, the media, or on the streets— will take him over the edge, yet again. As long as we can’t openly and civilly discuss important economic and social issues, we’re going to have silent supporters and loud surprises.