After Brexit and Trump, trust an issue with French polls

People set up voting booths at a polling station in a nursery school in Montreuil, outside Paris, on May 6, 2017.
People set up voting booths at a polling station in a nursery school in Montreuil, outside Paris, on May 6, 2017.PHOTO: AFP

NEW YORK (NYTIMES) - The mood in France is "tense." Europe is on "tenterhooks." The world is "on edge." The election in France on Sunday (May 7), pitting an independent centrist against a far-right nationalist, could signal the end of the "European experiment," lead to a "Frexit" and decidedly ring in a new era of reactionary populism and isolationism around the world.

All that stands between the devil we know and the devil we don't, the British and American news media have said frequently in recent days, are the polls predicting the outcome of the French election.

And so, again, we find ourselves awaiting an important election and poring over the numerical tea leaves.

But can we really trust polls? Remember what they said about the likelihood of a President Donald Trump? Remember Brexit? Remember arguing about pollsters who relied too much on calls to voters with landlines? Remember screaming into the void: "Who still has a landline?"

There is, however, a big difference between the poll results in France now and those before the US presidential election or the British referendum on whether to leave the European Union: The size of the lead.

Emmanuel Macron, the pro-Europe centrist, is ahead of Marine Le Pen, the nationalist who has vowed to engineer France's exit from the European Union, by more than 20 points in the latest French surveys.

In the most recent Ipsos poll, taken after the televised debate between the candidates Wednesday, Macron was at 61.5 per cent and Le Pen at 38.5 per cent.

In the second rounds of past presidential elections, French pollsters have, on average, been within 1 per cent of the actual result. And that, according to Claire Durand, a professor of sociology at the University of Montreal and president of the World Association for Public Opinion Research, can mean only one thing.

"There is absolutely no way that Le Pen could win," she said. "If it were to happen, it would be a massacre." In order for Le Pen to eke out an upset victory Sunday, millions of voters who have consistently signalled a preference for Macron since January would all have to change their minds at the last minute.

Polls have been wrong before, of course, but rarely are they that far wrong.

By comparison, in the days immediately before the US presidential election last year, Hillary Clinton's lead over Donald Trump in the polls was about 3 percentage points, or as FiveThirtyEight put it on Nov 4, four days before the election, "Trump is just a normal polling error behind Clinton."

The anxiety about a Le Pen victory is a function not only of the perceived inaccuracies of foreign polls, but the accuracy of French polling.

Le Pen received 21.3 per cent of the vote to Macron's 23 per cent in the first round of voting last month, just as the polls predicted. The fact that Le Pen, a firebrand accused of xenophobia and anti-Semitism, was able to beat nine other candidates and win enough votes to qualify for this weekend's runoff prompted concern among establishment observers that she could continue the trend of a global shift to the right.

For now, however, populism does not seem to be sweeping Europe or upending the status quo. Europe was similarly on edge before votes in Austria and the Netherlands, where nationalist candidates won in early rounds of voting but lost in the end - just as the polls predicted.

While the news media are apt to blame pollsters for getting important elections wrong, the pollsters are just as likely to blame the news media.

Many pollsters, including Durand, insist that the news media do not adequately explain the margin of error to voters, giving the impression that even a single-digit lead represents an inevitable victory.

"Without an explanation," she said, "those numbers are meaningless."