Despite improved communication technologies and other technical advances, traditional polling continually fails to accurately measure public sentiment or predict election outcomes. Polling fails because it ignores the most critical element that influences voting: identity. In the past year, we’ve been left with our jaws on the floor after surprise election results in the United States, the United Kingdom, Israel, and Colombia – just to name a few. Pollsters around the globe are in serious trouble, as they continue to rely on an outdated approach that is quickly losing legitimacy.

In an article published in Bold, ENODO Global introduced a new concept to help campaigns better understand voter behavior called ‘situational identity.’ It is unique to each individual and originates from a series of factors including groupthink, dominant public opinion, and a person’s worldview and social and economic standing in society. It is largely ignored by most pollsters, but can accurately predict an individual’s behavior and be applied to different demographic groups or across entire populations. Situational identity measures public sentiment concerning topics of interest, which helps us to better understand where exactly pollsters are going wrong. Most importantly, it provides insights into how to engage constituents to win elections.

It is no secret that the media plays a powerful role in shaping public opinion. The proliferation of left-leaning and right-leaning news sources has replaced “news” and flooded public discourse with sensationalized information. Individuals, activist groups, and media outlets highlight stories that reinforce their pre-determined opinions, support their organizations’ agendas, or increase viewership and ratings. Even those that report factually on politics often interpret and present stories in an emotionally charged way, which create biases and stigmas against others who support certain beliefs, candidates or parties. Moreover, the 24-hour news cycle of politically-slanted reporting creates and reinforces stigmas that carry significant implications for polling and elections.

For example, during the U.S. presidential election, many media organizations described then-candidate Donald Trump as a ‘racist’ and ‘sexist’. Left-leaning media organizations vilified Trump, subsequently creating a stigma that discouraged people from expressing support for him in different public settings and on social media. Out of fear of being labeled a “deplorable,” or being otherwise judged, many Americans who sided with Trump did so silently.

Up to election night, polls consistently put Hillary Clinton ahead by significant margins (as high as 5-7 percentage points). However, in a blog published the month before the election, ENODO predicted a Trump victory and credited the polling failure to this stigma gap – the silenced voters that pollsters missed. They never accounted for the situational identity of voters, who out of shame, fear, or embarrassment, refrained from publicly supporting Trump, but ultimately cast votes for him in the privacy of the ballot box.

The stigma gap took center stage again during this year’s Georgia House Race. Polls consistently projected that Democratic candidate Jon Ossoff would hold onto a slight lead over Republican Karen Handel. Again, public stigma influenced voters. After media outlets continuously highlighted offensive tweets, events, and perceived international faux pas during President Trump’s first few months in office, voters did not feel comfortable outwardly supporting a Republican candidate, even if they planned on voting for one. Subsequently, voters under-expressed their support for the Republican candidate and pollsters failed to account for the gap. Ultimately, Republican Karen Handel won the election by almost 3 percentage points.

The stigma gap also played a major role in Israeli Prime Minister Benjamin Netanyahu’s 2015 reelection. Throughout the Israeli election, Netanyahu was highly stigmatized by news outlets and activists alike who claimed his campaign featured racist undertones towards Palestinians. Much like the 2017 U.S. election, the media’s attack on Netanyahu discouraged voters from voicing their honest political views in public settings, which skewed polling accuracy. Netanyahu’s Likud party won a fourth of all available parliamentary seats, or thirty seats in total – at least nine more than expected.

Polling inaccuracies plague not only government elections, but also referendums on national decisions. Take the United Kingdom’s 2016 EU Referendum. Less than a third of polls predicted Brexit. By some accounts, there was an 80% probability that the country would vote to remain in the European Union, yet votes totaled in at 52:48 to leave.

Even Colombia’s 2016 Peace Referendum featured inaccurate polling related to the stigma gap. Pollsters projected support for the deal would be between 60 and 66%. They were so certain the deal would be passed that the ceremony to sign it into law was held the day before votes were cast. The next day, 50.2% voted against it. Perhaps the signing ceremony angered voters and encouraged them to vote to block the Peace Deal or perhaps the “no” camp had always been stronger than anticipated. They’d been attacked on social media for months by Peace Deal supporters and labeled “enemies of peace” by major news outlets – left too stigmatized to admit how they’d be casting their votes.

After the unexpected results of the Colombian Peace Deal Referendum, pundits suggested the inclement weather had played a role and discouraged some supporters from going out to vote. Why brave the rain if the deal was already so sure to pass? Similar theories have been postulated to explain Brexit and the last U.S. presidential election. However, proactive pollsters are advocating a technical change, combining landlines, cellphones and even online surveys to improve accuracy. Unfortunately, none of these options address situational identity and the impact of stigmatization on vote casting.

Traditional polling methods used to predict election results have proven to be inadequate time and time again in today’s information environment. This is not only a question of the loss of legitimacy surrounding polling. Inaccurate polling misleads the public and misallocates campaign resources. It represents the failure to listen to public opinion and reflects the extent to which countries’ institutions are out of touch with their populations – the problem that demands immediate attention.