Arie Kapteyn and Robert Cahaly, two pollsters among the few who anticipated the possibility of President Trump’s victory in 2016, warn against relying on polls that claim a Joe Biden victory in the next election, arguing that there is a large mass of President Trump voters characterized as “shy and hidden” who do not report their voting intentions in the polls.

Politico discussed the upcoming election with both pollsters. Both believe that the polls may be underestimating support for President Trump again, as they did in 2016. The reason is President Trump’s “shy” voters: people reluctant to share their opinions for fear of being judged.

“There’s a lot of hidden Trump votes out there,” they stated. “Will Biden win the popular vote? Probably. I’m not even debating that. But I think Trump is likely to have an Electoral College victory.”

Kapteyn, who directs the Dornsife Center for Economic and Social Research at the University of Southern California, commented that given this reality he decided to incorporate in his surveys a series of questions that include the “social circle” of the respondent, in this way he seeks to obtain information from people who may not be willing to speak for themselves but who are willing to speak for those close to them.

“We actually get a 10-point lead, nationally, for Biden over Trump” when asking voters who they personally plan to support, said Kapteyn. “But if you look at the ‘social-circle’ question, Biden only gets like a 5- or 6-point lead. … In general—and certainly on the phone—people may still be a little hesitant to say that they’re Trump voters.”

Many of the Trafalgar Group polls released last week show that Trump and Biden are statistically tied in key battleground states. Cahaly believes that the magnitude of President Trump’s “shy” voters could be even greater this year due to rising tensions in all quarters, particularly against his supporters.

Renowned political scientist Helmut Norpoth, who uses a particular form of statistical measurement to forecast election results, guarantees a broad victory for President Trump, predicting a 91% probability that the president will be re-elected.

Norpoth bases his claim on what he called the Primary Model, a statistical representation of U.S. presidential elections based on data dating back more than a century. As such, this forecast is unconditional and therefore not subject to periodic updates.

The model was correct in 25 of the 27 past elections. Even in the last election when the vast majority of polls announced a resounding defeat for Trump, Norpoth did not hesitate to be one of the few to predict a Republican victory.

His model qualifies incumbent presidents, the state of the economy, voter analysis, past re-election statistics, and alternation in power.