Donald Trump was just a normal polling error behind Hillary Clinton on the eve of the 2016 presidential election. The far-right National Front candidate in France, Marine Le Pen, by contrast, is an enormous, historic polling error behind the centrist En Marche! candidate Emmanuel Macron.
Macron won the first round of the French presidential election in April, 24 percent to 21 percent, and as expected, he has consolidated most of the support that went to candidates who did not make the runoff. Now, he heads into the runoff on Sunday with a huge lead over Le Pen. In an average of surveys conducted over the last two weeks, Macron has earned 61 percent to Le Pen’s 39 percent.
To win, Le Pen needs the polls to be way off. That’s possible. Clinton, for example, led Bernie Sanders in polls of the Michigan Democratic primary by 21 percentage points before Sanders’s shocking win there. But that upset was one of the biggest in U.S. presidential primary history. Polling misses of that magnitude don’t happen very often.
French presidential election polls in the runoff round have been fairly accurate over the past 50 years. In elections since 1969,1 an average of polls over the final two weeks of the runoff campaign (i.e., polls taken after the first round) has never been off by more than 8.4 percentage points.
|MARGIN BETWEEN RUNOFF CANDIDATES|
|YEAR||TWO WEEKS BEFORE ELECTION||ACTUAL RESULT||GAP|
The average error has been 3.9 percentage points, according to data collected by political scientists Will Jennings and Christopher Wlezien and aggregated by The Crosstab’s G. Elliott Morris. That means Le Pen needs an error of over five times the average in order to win. Now, we only have polling for the eight previous French presidential elections, so the average error may understate the possibility of a massive polling miss in France. Even taking into account the small sample size, though, the true margin of error in French polls is probably somewhere between 10 and 12 percentage points in either direction.
Le Pen’s situation is far different than Trump’s was in November 2016. Clinton was ahead of Trump in national polls by about 4 percentage points in the final FiveThirtyEight adjusted polling average. The national polls in the U.S. ended up being off by only 2 percentage points, or equal to the average error in previous presidential campaigns over the past 50 years. State polling was slightly worse, but even in the biggest shocker (Wisconsin), the polling average was off by 6 points. A polling miss of that magnitude in France would still result in an easy win for Macron.
But perhaps the biggest difference between the French campaign and the U.S. race is that there has never been a sign that a plurality of voters could be persuaded to vote for Le Pen in the runoff. Macron’s advantage over Le Pen has been fairly steady over the course of the year. Macron’s current lead, 22 percentage points, is well within the range of his average, 20 to 30 points, since January. In the U.S., Trump closed to within a couple percentage points of Clinton on a number of occasions. That showed that in the right moment he could come close to if not beat Clinton.
When Trump closed the gap on Clinton after the “Comey letter” late in the campaign, for instance, he was probably winning back a number of voters who had favored him before. Le Pen would likely need to win over voters who have never previously thought of voting for her (and even that might not be enough).
Indeed, Le Pen’s best shot is probably a combination of two polling mistakes or more. First, she needs the polls to be overestimating how many people will turn out — there needs to be a large number of deliberate abstentions (people casting a blank or spoiled ballot) or people not voting at all. Anywhere from 20 to 40 percent of the people initially supporting the other major candidates in round one have said they will not cast a ballot for Le Pen or Macron in round two. If that percentage went up dramatically, it would likely benefit Le Pen because most of the supporters for the other round one candidates are backing Macron. Second, Le Pen would need some unforeseen shift in preferences from Macron to herself. Voters telling pollsters they’re backing Macron will need to cast a ballot for Le Pen instead. (No such error occurred in the first round.)
If both of these types of errors occurred, it’s conceivable Le Pen could pull off a shocking victory. It took a similar multitude of polling errors in Michigan for Sanders to win. Much more likely, the polls are going to be at least somewhat close to the truth and Macron will be the next president of France.
Neil Paine contributed research.