I know, I know, I know… most of the media is telling you otherwise. They claim that the U.S. campaign is over. The only thing that they, the journalists, need to do is plan their dresses and tuxedos for the Clinton inauguration ball. Everything else is a formality.

The evidence they are using is poll data that indicates Clinton has an overall lead in the popular vote, the Electoral College votes are slanted towards her and that in a number of key states (Florida, Ohio, North Carolina) she has a 3 – 5 % lead. The excellent www.fivethirtyeight.com polling site puts her chances of occupying the White House at about 88.4%.

They might be right, but here are a couple of points on why things could change very drastically in this election.

First, all polls have a margin of error of roughly 4%, so unless the lead is above 5% they are essentially saying the overall vote is too close to call. This is the reason why you currently have a Monmouth poll predicting a electoral victory for Clinton of 12%, while at exactly the same time the Los Angeles Times/USC Tracking poll is showing a 1% lead for Trump.

Too Embarrassed to Admit It in Public

More importantly, some of these same pollsters have been spectacularly wrong about two recent major British votes: the Tory majority victory in the UK election (May 2015) and the Brexit Referendum (June 2016). There are similar circumstances around those votes in Britain and the upcoming U.S. election.

The most obvious similarity is that they were bitter campaigns that showed the country was deeply divided. This meant that there was a strong social punishment for admitting in public to voting for the ‘wrong’ side.

After the Brexit vote, UK talk radio was full of callers who asked how to deal with their friends and family who had voted for the other side. People were seriously asking if they should break up friendships, get divorces or disinherit relatives for voting for the Leave/Remain side.

Today, the same circumstances exist in the U.S. It would take a brave woman in northern California to admit to her liberal friends that she was going to ignore Trump’s alleged sexual misconduct and vote for him because she liked his stance on illegal immigration. A New Yorker supporter of de Blasio would have equal difficulty admitting in public that they preferred Trump’s ‘get-tough-on-crime’ stance to Clinton’s.

Trump is not Romney

This is a fundamentally different contest than the Obama vs. Romney election of 2012. Very few of Romney voters were embarrassed to admit that they supported him in public: while, potentially, many of Trump’s supporters may be afraid to do so.

The phenomena is called “social desirability bias’ (or more commonly ‘The Bradley Effect’ after Tom Bradley the 1982 California Gubernatorial candidate who was surprisingly defeated by racist voters too frightened to admit their prejudice to pollsters) and it is particularly strong in this age of anti-establishmentism.

It is the reason why the UK pollsters failed to accurately predict the results of those recent votes. They did not factor in the small, but significant group, of the electorate (2-4%) who were going to vote for the unpopular candidate or choice, but were too embarrassed to admit. These voters will, in public or to a pollster, nod or say what their friends expect them to say: but, in the privacy of the voting booth do something else.

None of the American polls factor “social desirability bias” into their analysis. They could not do so, as it is almost impossible to accurately measure. However, it may explain why Trump does better in online polls where voters do not have to reveal to another person who they will vote for.

“A week is a long time in politics”

There are three weeks to go until election day, time enough for any number of surprises to emerge. There will be more hacked e-mail scandals to come. There will be another major televised debate on Wednesday. Anything may happen that could change this election.

Finally, Clinton’s apparent lead in the polls has not been propelled by support for her or her policies but out of dislike for Trump’s objectionable comments about women and the numerous accusations of sexual harassment against him. This means that Clinton’s support is fundamentally weak and can be reversed.

So to be clear, I do not support either Donald Trump or Hilary Clinton. That is not my job as a journalist. And, Clinton may still win the election, it is just that her path to the White House is not as clear as many observers are telling you.