Can Le Pen Win? Yes. Will She Win? No.

Tweet

On the eve of the French presidential election, an average of polls puts Emmanuel Macron ahead of Marine Le Pen, 61% to 39%. That's a massive gap.

To give some perspective, in Donald Trump's "upset" win in November, he beat the polls by only 2% in the popular vote. Le Pen will have to beat them by ten times that.

Nothing like that has ever happened in modern French politics. In the last eight presidential elections, the average polling error has been 4%, and the most the polls have ever been off was 8.4%. That was in 2002, when Jacques Chirac beat Marine's father, Jean-Marie Le Pen, 82% to 18%, a victory margin substantially larger then the polls had predicted.

An interesting article at FiveThirtyEight, points out that errors of 20% or more do happen occasionally. When Bernie Sanders upset Hillary Clinton in the Michigan Democratic Primary, he beat the polls by 22%. Interestingly, that's exactly what Le Pen needs to do.

But such examples are extremely rare.

To do that, Le Pen will need a much higher number than predicted of deliberate abstentions - people who did not vote for Le Pen in the first round but refuse to cast a vote for Macron - along with getting an almost unheard of proportion of undecided voters to go her way and/or people doing the opposite of what they told pollsters. That this didn't happen in the first round - if anything, Le Pen slightly underperformed polls and expectations - does not bode well for Le Pen.

By almost all accounts, Le Pen did badly in the final debate, looking weirdly smug against a personable Macron. A last minute "dump" of hacked emails from the Macron campaign, allegedly containing all sorts of negative information - some of which the Macron campaign has declared to be falsified - came too late to matter, and may have even stirred up sympathy for Macron. It's essentially been embargoed by the French media, in any case. And, as of this writing, France is entering a twenty-four hour period in which all discussion and reporting on the election is prohibited by law.

Curiously, the prediction markets still give Le Pen a 15% or so chance of winning, a probability much higher than what her low standing in the polls should justify. The markets seem to know something that we don't, but what that is is unclear. It translates into Le Pen having an outside chance, not no chance but an outside chance - a chance equivalent to rolling a one on a six-sided die.

One in six.

Stranger things have happened.

But usually they don't.