Unless she can pull of the biggest upset in France electoral history, Marine Le Pen is going to lose this Sunday’s vote for president of France to Emmanuel Macron. That has been true every day since April 18th of this year, when both party leaders finalized their candidacy for president. It was even true back on January 4th, when we started tracking this race. And it will be true on Sunday when voters go to the polls and (almost certainly) elect Macron, centrist candidate for En Marche!, as their nation’s next leader.
Two weeks ago, Emmanuel Macron bested Marine Le Pen in the first round of the election 24% to 21%. Way back then, Macron was only 26 points ahead of Le Pen in public opinion polls and had a 97.5% probability of victory. It was then that I wrote the following:
Ultimately, it’s better to trust polling than punditry on this one. Public opinion numbers have proven accurate in previous French contests, and even developing a model that is bullish on Le Pen’s chances of victory tells us that 49 times out of 50, Emmanuel Macron becomes the next President of France. Thus, the question pundits should be asking is not “can polls err enough to elect Le Pen?” but “can Le Pen pull off one of the biggest upsets in electoral history, anywhere?”
Have things changed sine April 24th? As it turns out, yes, Le Pen may have gained some ground — and then lost it. Although she sat at “just” a 19.5% polling deficit at the start of May, Macron has since increased his lead to 22%. The chance of Le Pen winning has decreased accordingly, topping out at 3.8% and now 3.1%.
The question remains, can Le Pen beat those odds? Can she turn that 22% deficit into a win? It is possible. But very, very unlikely.
Primarily, the statistical model I built to forecast this year’s election says that Le Pen has almost no chance of winning. This is mostly the case because we have never seen an error in France election polls large enough for her to be victorious.
|Year||Poll Margin (%)||Election Margin (%)||Error (%)|
According to data from professors of political science Will Jennings (at Southampton University) and Christopher Wlezien (at the University of Texas at Austin) the average error in the polled margin of victory between the candidates and actual margin of victory has been 3.9%. Those polls have been wrong just once in the eight elections since 1969. The largest polling error (so far) was in 2002, when Gaulist Jacques Chirac beat Marine Le Pen’s father, Jean-Marie Le Pen by 8.4% more than expected.
With those numbers, we can estimate that polls should land within 15% of our estimates 99% of the time. 95% of the time, error will be less than 12%. In France, an error big enough to elect Le Pen almost never happens. They also don’t happen elsewhere; in #Brexit, for example, the polls were off by just 4%. Donald Trump won the US election by overperforming expectations by just 1-2%. A 22% upset just does not happen but once or twice in a lifetime.
However, those are just the hard numbers based on historical polling error. What happens if we, like my model, also adjust the estimate for a very large number of undecided/abstaining voters?
That is the second step in explaining why Macron is such a favorite.
According to the polls, a quarter voters in France are undecided or abstaining (heretofore referred to just and undecided) from voting in the election. Based on polls of the losing round one candidates, most of those undecideds are disenchanted remnants of far-left candidate Jean-Luc Mélenchon. The model assumes that, like the first round, anywhere from 5 to 10 percent of those undecideds will actually vote. Then, it gives the votes proportionally to Le Pen and Macron, often giving an edge to Le Pen to produce a more optimistic forecast for her (than, say, The Economist’s).
But, what would happen if Le Pen won a majority of undecideds? Maybe even all of them? That’s not too hard to figure out. All we have to do is add the undecideds to whoever we think will win them.
Because France election polls adjust poll numbers for candidates to be a percentage of only those who are voting, we first have to readjust Macron and Le Pen’s numbers. Changing the numbers from that proportion of voters to a proportion of the voters plus an (optimistic) 10 percent undecideds yield the following adjusted polling numbers:
Then, its simply a case of allocation. if Macron gets a little more than half of the undecideds, like his numbers show, he will end up with 60% to Le Pen’s 39% (duh, that’s just reversing the math we just did!). But, if Macron gets, say, 30% of the late-deciding voters (as Clinton did in Wisconsin in 2016 US election) he may earn just 57% of the final vote to Le Pen’s 42%. Then, one of those once-in-a-lifetime 15% polling errors could elect Le Pen. Even if Le Pen wins all of the undecided voters, she still needs a 10% polling error to win the election.
There are two plausible scenarios, in my mind, for how Le Pen could earn the support of more undecideds than most people think. First, and most obviously, she could have a late surge among the French citizenry that either causes voters to turn out for her or adjusts the preferences of voters who could have supported either one. Secondly, Le Pen could hope to depress turnout among ex-Hamon and Mélenchon voters — since they disproportionately support Macron, their abstention would help Le Pen.The probability of Le Pen earning enough undecided voters (plus getting lucky elsewhere) AND having a large enough polling error to win, says my model, is just 3.1%.
But, that is a best case scenario for Le Pen.
Put simply, it’s really, really, really hard to imagine enough stars align to elect Marine Le Pen. As I tweeted the other day, it’s Macron’s election to lose.
This all being said, the aim of building a forecast model is to think probabilistically and convey a wide range of outcomes. To that effect, this vote histogram is useful:
The graph above shows all of the possible vote shares for Emmanuel Macron, and how likely they are to occur. Macron’s vote share is on the x axis, and the y axis is the frequency with which we can expect that outcome to occur.
Effectively, there is a 88% chance that Macron will earn more than 55% of the vote.
|Macron Vote||Chance of Occurring|
|61% (or more)||50%|
This all to say, don’t be surprised if Macron doesn’t hit the nail exactly on the head come Sunday. This is especially true given that no polls will be released after Friday at midnight. Given that the polls show a surge against Le Pen, it may not be too unreasonable to think she might underperform polls.
But at the end of the day, the forecast is what the model says it is, and it says that Le Pen is way behind.
Here is a list of things more likely than a Le Pen victory:
This all to say that Le Pen will have a hard time becoming France’s next leader. It could happen, of course, but your money is 33 times safer on Emmanuel Macron.
Speaking of money, what do other sources of predictions say? As some experienced viewers of elections will know, betting markets frequently make predictions on these contests. According to the statistical model I built for the France election — which does everything I’ve discussed above, and more — Marine Le Pen has a 3% chance of winning the France election. Other math-based modelers like The Economist rate Le Pen’s chances at < 1%. The betting markets, however, have Le Pen at 10-12% and Ian Bremmer, of risk consulting firm Eurasia Group, puts the chance of a Le Pen victory at 30 percent (his model, however, is just based on “externalities” that could elect Le Pen, not hard numbers).
There are a few reasons to think that these predictions based on market prices and conventional wisdom will be wrong on Sunday. In short, (1.) claims of a “shy” Le Pen vote aren’t grounded in evidence; (2.) if round one polls are any indication, polls in round two will not err by anything more than 1-2%; (3.) the winners in round one of France election shave gone on to win round two 70% of the time; and (4.) the model takes into account a healthy amount of uncertainty, and even then, Macron is about as heavy a favorite as they come. For more, read what I said in part three of this series.
Another way to think about the forecast is in odds. IE: I say Le Pen wins once for every 33 times Emmanuel Macron wins the presidency. The Economist says that may be closer to 1 in 100 or 200, which seems a little too low, all things considered. Betting markets say she gets a 1 in 7-8 shot at victory (which is higher than some of them rated Donald Trump’s chances!) and Ian Bremmer says Le Pen will win once for each time Macron wins twice. That is just way too high when you’re down 22% in the polls.
On Sunday, May 7th voters in France will pick their new president. According to hard math based on public opinion polls, Marine Le Pen has almost no chance (just 3%) of winning. Although we will very likely be calling Macron the president soon, some may be wondering whether his party (or lack thereof) will be able to win a productive number of seats in the French Parliament after those elections on June 11. And who will Macron pick as Prime Minister?
The answer to those questions and more in the fifth and final installment in these series of posts, coming after Macron (probably) wins it all at, which you can find out when results roll in at 1:00 or 2:00 EST on Sunday. Tune in to my twitter updates throughout the weekend.