U.S. pollsters got a lot of heat last year for having failed to predict Trump’s win. To begin with, polls are never a prediction. Second, as we now know, the nationwide polls have actually been quite accurate. In the surveys right before the election, Clinton was leading by 3%-4%. On Election Day, she won the popular vote by 2%. The problem is that a U.S. presidential election is not a nationwide election. It’s a state-by-state election and therefore it doesn’t only matter how many votes a candidate gets, but where he gets them!
Harry Enten from fivethirtyeight made a good point in that respect the other day: The recent French presidential election will hardly be remembered as a failure for the polling industry, when in fact, pre-election polls were quite off. The average poll conducted after the first round saw Macron ahead by 22%. On Election Day, he won by 32% over Le Pen. That’s a gap of 10% between the surveys and the actual result, and the biggest one in recent French polling history.
Guy Verhofstadt, the former Belgian Prime Minister, made a similar point in the New York Times recently. There was much talk after the French elections that Macron only won because voters voted against Le Pen and not for him, and that therefore, he would have little of a mandate. Trump’s win, on the other hand, was often reported as a fundamental change in politics. Well, as Verhofstadt argues, there is another way to look at it which is actually based on numbers: Trump won with 46.4% of the vote and a voter turnout of 55%. Macron won with 66% of the vote and a turnout of 74%.