Never before in modern history has a French presidential election been punctuated by so many unforeseen events of all kinds, judicial and electoral. It ended up on the April 23 first-round vote with a four-way split, ranking centrist Emmanuel Macron first with 24.01%, followed by Marine Le Pen of the Front National (FN) on 21.30%. François Fillon of Les Républicains was on 20.01% and Jean-Luc Mélenchon of the far left on 19.58%.
Macron and Le Pen are therefore heading for a second round run-off on May 7. First polls gave a ratio around 60/40 in favor of Macron.
As in recent campaigns in democratic countries, the French public has been deluged with polls of various kinds. Each shift in voters’ stated intentions is passed under the microscope and commented upon. While the very reliability of polls has been questioned given the unexpected victories of the Brexit vote in the UK and Trump in the US, pollsters regained legitimacy with their rather precise predictions of the first round result.
The central question is how predictable the final outcome can be based on an extrapolation of the polls. However, if using probabilities can capture the appreciation of what may happen, in the case of a single event—a presidential election, say—the notion of probability doesn’t actually say anything about what is really going to happen.
To make a credible prediction, we need a model that allows us to consider the evolution over time of the different voting intentions and so to predict what they will be on the day of voting. I have developed a model using what’s called “opinion dynamics” within sociophysics, a growing field of research especially among physicists. This model allowed me to predict the victory of Donald Trump as early as the summer of 2016. Applying the model to the French presidential election shows that Le Pen is out of reach of victory with an active glass ceiling which maintains her below 50% of people’s voting intentions.
Nevertheless, beyond the declared voting intentions for the two candidates in the second round of the French election, a phenomenon that’s more difficult to gauge is the abstention rate. As I shall demonstrate, if people decide not to vote, it could tip the balance in favour of the candidate placed in second place in the polls.
Polls currently suggest that the abstention rate in the second round could be the highest ever for a French presidential contest—between 19% and 22%. In general, abstentions benefit candidates from outside mainstream political parties, but this time both of the mainstream parties were eliminated in the first round—Le Pen and Macron are thus both outsiders.
This is why it’s essential to have a general study of two competing candidates in an election and then apply it to a specific case. I have gone into the details in a scientific article, but the implications can be explained relatively simply.
In the runoff vote between Macron and Le Pen, her voters will largely be those who want to vote for her, something that won’t be the case for a percentage of those who vote for Macron. When Le Pen’s father Jean-Marie, founder of the FN party, unexpectedly made it into the second round of the 2002 presidential election, the country’s other parties formed a “Republican front”—collectively voting against the far-right FN. In the first round, the leading candidate was Jacques Chirac with just 19.88%, followed by Jean-Marie Le Pen with 16.86%; in the second round, Chirac won with 82.2% over Le Pen’s 17.8%. That required a large number of voters who’d initially supported other candidates to vote for someone who wasn’t their first choice.
However, the second round of the 2017 election is unique as a large number of voters who do not want the FN to win will have to vote for Macron whose politics they strongly oppose. Trapped between a repulsion for Le Pen and an aversion for Macron, they will force themselves to make an unpleasant choice and vote against the far-right. Nevertheless, to implement this choice will require them to cast a ballot for Macron, which for some will be like swallowing a large, bitter pill.
That’s why it’s quite likely that a significant number will take advantage of any “good excuse” at the last moment to not take the time to do so, and so boost the abstention rate. I define this novel behaviour as an “unconfessed abstention”, which in turn will produce a differentiated abstention at the collective level. This means we can realistically assume that there will be more abstentions among those who say they will vote for Macron than among Le Pen supporters. On this basis I can calculate the difference between the rates necessary to make up the distance between the two candidates and so reverse the final ranking.
For example, if Macron were polling at 58% and Le Pen at 42%, and 90% of those who say they’ll vote for her actually do so, she would get more than 50% of the vote if the participation for Macron falls below 65.17%. Or, if only 65% of those who say they’ll vote for Macron actually do so, Le Pen would win with 50.07% of the vote.
If the polls predicted 55% Macron and 45% Le Pen, and 85% of those who say they’ll support her do so, she would win with more than 50% of the vote if fewer than 69.55% of Macron’s supporters actually vote for him. Or, if only 69.5% of those who said they’d vote for him do so, Le Pen would win with 50.02% of the overall vote.
By definition the unconfessed abstention rate cannot be evaluated by polls, making it more difficult to forecast the final outcome. This means that in addition to the battle to reach a maximum of voters in the polls, mobilisation on the last day is likely to be decisive. So, despite the “Republican front” and Le Pen’s polling below 50%, she could still become the next president of France if the differential abstention rates are in her favour—and it wouldn’t take much for that to be the case.