The Big Data calls for an unexpected winner of the first round
Big Data analysis is on its way to revolutionise the way we predict political elections.With all the social networks, there is plenty of data available to analyse new trends. Half of the French population is on Facebook and Twitter has 15 million users in France.
Some companies have understood it and they use it to predict. They are three main “Big Data specialists” in France who already won their spurs. As an example, they have predicted Fillon’s victory in the first round of the primaries but also the election of Donald Trump and even the Brexit. Now predictions from Filteris, Vigiglobe and the algorithm Predict My President are not calling for a duel between Le Pen and Macron, but for another candidate to win.
Another Big Data specialist is Google with its service Google Trends. Once again the results are different from the polls, but also from the other Big Data analysts.
Traditional polls have lost the public’s trust in the previous year
It was Brexit first and then Donald Trump. These two major events made an upset partly because they were both wrongly predicted by national polls. If we look at the current French presidential election, opinion polls made similar mistakes during the primaries of Les Républicains giving Alain Juppé winner of the first round with Nicolas Sarkozy followed by a victory of Bordeaux’s mayor in the second round.
The latest polls from the first round have pretty similar results with Emmanuel Macron first followed by Marine Le Pen to win. Then Jean-Luc Mélenchon and François Fillon are joint only a few points behind. This tendency has not changed in the latest weeks.
The main weakness of these traditional polls is how they operate. They base their results on a certain sample representing the population. But it is only composed of a few thousand potential voters only. It is this lack of consequent data that can make these polls wrong.
Therefore, let’s have a closer look at these new analysts.
Filteris: Fillon and Le Pen for the Second round
I.D: This Canadian company was created in 2002 has created BuzzPol in 2007. It analyses the buzz generated on the web and on social networks in an electoral context. It differs from the traditional polls as it analyses the perceptions, thoughts and opinions expressed on the web on a large scale.
Its algorithm looks at all these angles quantitatively (who provokes the more reactions?) and qualitatively (Which tonality?) and gives a score out of 100 redistributed between the candidates.
Prognostic: In its latest measure, Le Pen leads with 22.72%. He is followed by Fillon (21.57%) then Mélenchon (21.34%) and finally, Macron with 20.03%. A bit more than two points separates the four candidates.
It is the 16th analysis made by Filteris since the 3 April. The first 14 ones have predicted a duel between Fillon/.Le Pen in the second round. Only the 15th measure, on Thursday predicted the couple Mélenchon/Le Pen.
Can we trust Filteris? The company has successfully predicted Trump’s victory and did not give an analysis on the Brexit. It has also predicted Fillon’s victory in the primaries. However, they were waiting for Sarkozy in the second round against its former prime minister instead of Juppé. Filteris also made a mistake during the socialist primaries. On the first round, it gave Hamon third instead of first.
Vigiglobe: Fillon probably qualified, Macron and Mélenchon following closely and Le Pen out of the game.
ID: Vigiglobe is a French start up lead by Leendert de Voogd, the former world director of the polling institute TNS. The startup has developed analytic tools for elections to find weak signals translating dynamics that polls do not see or see too late.
How do they work? Vigiglobe uses machine-learning algorithms (the machine learns thanks to an AI). Like Filteris, they analyse quantitatively and qualitatively messages posted on social networks. The startup has access to exclusive content thanks to partnerships with Facebook and Twitter.
Prognostic: Unlike its opponents, Vigiglobe does not risk the predictions game. However, on Thursday the start up stated that Fillon “have good chances to be qualified for the second round”. The algorithm which looks at the positive opinions is calling for a leading trio composed of Fillon first, followed by Macron and Mélenchon.
The main difference is that Vigiglobe thinks that Le Pen is seriously losing momentum and will not be qualified. However, the CEO de Voogd reminds that the attack on the Champs Elysées on Friday benefited Fillon and Le Pen compared to the previous days.
Can we trust Vigiglob? Vigiglob was true concerning the American election, the British referendum and both primaries.
Google Trends: Mélenchon and Fillon in the Second round
ID: Google Trends is a tool offered by the browser Google. It allows you to look at the number of searches on a certain subject. You can change several settings such as the country, the period, the category, the kind of web search, etc.
How does it work? Google Trends looks at all the searches on google concerning each candidate and gives a graph relating the evolution of each candidate popularity in a certain period. Each data point is divided by the total searches of the geography and time range it represents, to compare relative popularity. Otherwise, places with the most search volume would always be ranked highest. It results in a score out of 100.
Prognostic: If we look at the last 90 days, on Saturday, 22nd of April, Fillon is leading in average with 32% followed by Mélenchon with 26%. Marine Le Pen (13%) and E. Macron (10%) are far behind.
Credibility: Google trends predicted well the American elections as well as the favourites for the second round of the two latest French elections. However, the trends were not able to give the winner in both cases.
Predict My President: The couple Le Pen and Fillon — Fillon for the second round.
I.D: It’s a project led by five students from Telecom Paris Tech School. The objective is to overcome the polls’ failures by replacing opinion polls with a socio-economic analysis of the voters mixed with an analysis of the popularity of each candidate on the web.
How do they work? A software gathers all the information on Twitter and Google searches to determine each candidate’s popularity. All this information is crossed with a data bank of socio-economic statistics starting in 1981 (jobless rate per French department, results of local elections, etc.). They use governmental open-data platforms and the Insee (French national institute for statistics and economic studies). An algorithm weights the different analysis and results in giving vote predictions.
“When Big Data and socio-economic analysis replace polls”
Prognostic: On 19 April, Predict My President forecasted a duel Le Pen (24.13%) / Fillon (21.77%). Macron was third with 20.32% followed by Mélenchon with 18.66%.
As Predict My President is brand new in the game, we will see tomorrow if they are right.
“This is the most incertain French election that I can record and the most important” Dominique Moisi
To conclude, most of the Big Data analysts were naming the former prime minister winner of the first round and his opponents are uncertain. It is Le Pen and Macron first most of the time but also Mélenchon.
On the contrary, the latest traditional polls are still calling for a duel Le Pen/Macron for the second round with Mélenchon and Fillon behind but still very close to them.
However, this election is one of the tightest that France has ever known in the history of the Vth Republic and we cannot be certain of the results. This is comforted by Dominique Moisi, a French political expert, who stated during a speech at the Association of European Journalists: “This is the most uncertain French election that I can record and the most important”. We will only have the answer Sunday night…
