(Almost) Anything Could Happen in Round One of the France Presidential Election

Round one of the France 2017 election is way too close to call. I use our election forecast to show you why.

By G. Elliott Morris / April 22, 2017

 in European Politics

This is part two in a series on the 2017 France presidential election. Read part one here. Read part three here.

The first stage of the 2017 France presidential election occurs on Sunday when voters will choose their candidates for the second, head-to-head stage of the contest. That will occur so long as no candidate receives over 50% of the vote — which, according to polls, is close to impossible.

Indeed, by 1:00 PM EST on April 23, we should know a little more about who the next president of France will be — but what exactly do we know right now?

In short, the first round of the election is a mess of uncertainty, where no candidate has a clear lead. Our forecast of the election shows that no prediction can be made with certainty, and conventional wisdom about what might happen is even more lacking.

The long version of the story is as follows.

The Polls

If the polls are to be believed (more on that below) then the France election is a mess of vote intentions. In short, the candidates are simply too close for us to make an effective call in the election.

According to my weighted average of French public opinion polling, centrist candidate for En Marche! Emmanuel Macron currently leads the pack of other viable candidates. As of April 22 at 10:00 AM, Macron is earning the support of 23.7% of voters. He is roughly one point ahead of far-right candidate for the National Front Marine Le Pen, who sits at 22.4%.

Three percentage points behind Le Pen, at 19.5%, is center-right Republican François Fillon. Two months ago if I were writing this piece, I would have picked Fillon to become the next President of France, but due to a very public scandal that is no longer the case. But, it could still happen (again, more on that below).

Finally, in terms of viable candidates, half a percentage point below Fillon is the far-left candidate for Unsubmissive France Jean-Luc Mélenchon at 18.9%. Mélenchon has increased his stance in the polls nearly 10% in the past two months; that’s a very significant rise, with all sorts of implications.

These polls are not gospel, however, and there is reason to think they may not be correct.

Error in the Polls

Just as there is error in US election polls, there is error in France election polls. Average error in the first round of the past eight presidential elections (since 1969) is just over 3%, according to data from political scientists Will Jennings, of the University of Southhampton, and Christopher Wlezien of the University of Texas at Austin. That’s 1% more than the 2% average error for US election polls.

Year Poll Margin (%) Election Margin (%) Error (%)
Error in Round 1 France Election Polls
1969 16.00 21.16 5.16
1974 13.50 10.65 2.85
1981 -5.75 -2.47 3.28
1988 -11.50 -14.16 2.66
1995 3.50 -2.46 5.96
2002 6.28 3.02 3.26
2007 4.46 5.31 0.85
2012 -1.56 -1.45 0.11
Average     3.01%

The margin of error on these polling misses is just south of 4%.

This means that we can expect error as large as 7% to occur “most of the time” in France elections, sometimes reaching as high as 8-9%.

Of course, it’s possible that polls miss by even more than 9%, but I would not count on that happening (surely do not put money on a Benoit Hamon comeback).

It is certainly worth noting, however, that error in recent years has been less than one percent. So, while it’s possible that error in the polls is very large, it’s also possible that we have great polls that paint a nearly exact portrait of what will happen on Sunday. But that’s not useful; we want to know what may happen if something goes wrong, not if everything goes right.

If we look at that 3% error in context with the 2017 election, we see that an average miss would be enough to close the 3% gap between Marine Le Pen and François Fillon, putting the latter in the runoff with Emmanuel Macron. A miss of 5% (very possible, though not exactly probable) would be enough to put the surging Jean-Luc Mélenchon in the first place spot, primed for a stunning runoff with Macron.

So, taking into account this error, can we give some odds to the candidates? You betcha.

Forecasting the Election

In order to forecast the election, I developed a model to take into account the historical uncertainty in France public opinion polls, among other factors. Of course, the model generated a forecast) for today with the following probabilities:

Candidate| Make Round 2 (%)| Win Round 2 (%) |— Macron| 73.6| 70.1 Le Pen| 62.7| 4.0 Fillon| 34.4| 11.2 Mélenchon| 29.2| 14.1 Hamon| 0.1| 0.1

We see above that Emmanuel Macron and Marine Le Pen have about double the odds of making the second rounds that Fillon and Mélenchon have — but, Fillon and Mélenchon’s odds aren’t really bad at all. In fact, the candidate with lowest probability of making round two (Mélenchon) has roughly double the odds that Donald Trump had of winning the 2016 US election. On the other hand, the chance that Macron does not earn his first-place spot is just more than the chance of flipping heads/tails on a coin two times in a row. Try it; that’s pretty easy to do.

If someone told you that you would get in a car wreck one out of three times you went out driving, would you drive?

The forecast also has the ability to pick certain head-to-head matchups based on the above polling and error.

{:.colored} Matchup| Chance of Happening (%) |— Macron vs Le Pen | 40.5 Macron vs Fillon | 18 Macron vs Mélenchon| 15 Fillon vs Le Pen | 12 Mélenchon vs Le Pen| 10 Mélenchon vs Fillon|4

Breaking down those round-two probabilities into matchups makes us even less certain of tomorrow’s outcome. There are less than majority odds for any outcome.

In short, there is no bet on the France 2017 election that is remotely safe to make. The election, simply, is way too close to call.

What the forecast can’t do, however, is predict an outcome that has no basis in the data. In other words, if the polling is simply bad data, or an event causes a shift in the polling we have before the election, that won’t be reflected. I have some thoughts on those possible event.s.

There’s More Uncertainty than We Expect

It is of course possible that some other event may impact the result of the election in a manner that the polls don’t pick up. What if there is a “shy” Le Pen vote? What if pollsters are herding their numbers? Maybe there is a late surge for Mélenchon between the 21st, the last day to publish a poll, and election day (23rd)? Of course, there is also the possibility that the large (25-30%) undecided vote breaks more in favor of one of the candidates than the others. Let’s go issue-by-issue.

"Shy" Le Pen vote

The theory I hear the most about why polls will miss is that there is a faction of French voters that are lying to pollsters. In fact, I hear this the most about any country’s election. But is there any data to back this up?

The same “shy voter” theory emerged in the aftermath of the 2016 US election polling “miss” (which maybe wasn’t a miss after all) to explain why Trump over performed expectations. Well, there is no reason for us to think that that’s true. In other words, there’s no data that shows people are lying to pollsters — at least about candidate vote intention, as there are some questions about which survey respondents will will lie.

Nevertheless, analyses show that this “shy Trump” theory didn’t play out in the United States. We don’t really have a reason to think it will happen in France, either.

However, this is not to say that Le Pen would not be underestimated for other reasons. It has been hypothesized that she may pick up some of the last-minute deciders and vote-changers because of a recent terrorist attack in Paris. I think this makes the most sense for why Le Pen would over perform. After all, right-wing parties in Europe have a pretty bad record at beating expectations. Just recently, Geert Wilders’s Freedom Party had a bad showing in the Netherlands General Election.

So in the end, there’s not a clear evidence that makes me believe Le Pen would beat her numbers. That does not change the fact that it could still happen.

Poll herding

Recently there has been a lot of talk about herding in France opinion polling. “Herding” occurs when pollsters fudge their numbers, adjusting more to the current polling average as not to appear to deviate from what the others are saying. In an article by Harry Enten of FiveThirtyEight, he discusses why there may or may not be herding in France opinion polls. Even if pollsters aren’t faking their numbers, something is certainly going on to make the numbers much more consistent than we should expect statistically.

However, it’s impossible to know just how much this will influence our forecasts. Thus, increase your expectations of uncertainty for the election.


As the average shows, far-left candidate Jean-Luc Mélenchon recently experienced the largest surge I have ever seen in public opinion polling. Some people are saying this is cause to believe he will over-perform on election day. While this is of course possible, I’m not sure there’s any real correlation between a late surge — especially one that has leveled off, as Mélenchon’s has — and exceeding election day expectations. Another +1 for uncertainty, although Mélenchon would have to induce the third-largest polling error in French history to take first place. But, an average polling error would be enough to get him into the runoff with Emmanuel Macron.

Undecided voters

The biggest reason to worry that polls will be off in the France election is the number of undecided, or “abstention” voters According to polls, 25-30 percent of French voters are going to abstain from voting on Sunday. While technically this could mean they won’t vote, there is always the possibility that they will decide last minute to vote for a candidate. And, if a lot of voters decide to do so, they could swing the election.

In the past week, the number of undecided voters has decreased about 3%. It’s hard to tell which candidate received their support, but the most obvious candidate is Emmanuel Macron, who received a 1-2 percentage point bounce in the polls last week. It’s possible that he receives more, but it’s really impossible to say. That’s just another reason to add a point to our uncertain “index,” which is now at a +3.

There are simply a lot of reason why polls could be off, and nothing is safe to say.

Wrapping Up

We have learned in this article that the France election is, simply, a mess of polling. The candidates are all squished together in a five percentage point range from 23% to 18%, and an average error could turn the tables in any which direction. This begs the question: how big of an error should we expect to see?

We have gone over a few reasons why that error may be bigger than we expect, departing from a trend of increasing polling accuracy in the last two elections in France. For one thing, polls appear to be adjusting their numbers to that of either that past polling or everyone else’s, which really would mess up our forecast. Other considerations include the large amount of undecided voters and movement in the polls from Friday to Sunday, when polls aren’t allowed to be released in France (note: some will be released in overseas media markets).

In the end, however, we can’t make a call. Everything is simply too uncertain. It’s true that Emmanuel Macron and Marine Le Pen have the best chances of making the runoff, but we should expect another pair to make it 60% of the time.

In sum, don’t be surprised if your candidate exceeds/doesn’t meet expectations Sunday afternoon. Polls have been wrong before — with some “big” misses recently — and they can certainly be wrong again.

… . .

That’s all for our final forecast of the first round of the French Presidential election. If you haven’t read part 1 of this series, read it here. Part three, covering round 2 expectations, is coming soon!

In the meantime, tune in to my twitter for updates to the election forecast, polling, and more.





comments powered by Disqus