French Presidential Election of 2017

alt="France 2017 Election (first round) Vote Percentages">

The pie chart shows the official results of the first round of the 2017 election. [23 April, 37M voters (78% turnout), plurality voting, 2.6% invalid "spoiled" ballots.] The second round was won by Emmanuel Macron with 66.1% of the votes versus Marine Le Pen's 33.9%. [7 May, 31M voters (75% turnout), simple majority vote, 11.5% spoilage.]

But in addition to the official vote, pseudo-election exit poll experiments using approval and score voting were done by a team of academics with government funding and cooperation. And many professional pollsters (often funded by the media) carried out pre-election polls using various methods.

alt="Paris Match poll">

We begin with the Paris Match pre-election poll "among a sample of 1005 people representative of the French population aged≥18. The representativeness of the sample was ensured by the quota method (sex, age, occupation of head of household, level of education) after stratification by region and category of agglomeration. Telephone interviews 7-8 April 2017."

"Question: For each of the following politicians, tell me if you have an excellent opinion, a good opinion, a bad opinion, a very bad opinion or if you do not know enough to say?"

This really is best interpreted as a score-voting poll with 4 score levels (plus "don't know enough"). But unfortunately Paris Match only published the percentages of people expressing the two good opinions (lumped together). That is what we have called "approval" in the figure above. Also note, the Paris Match poll asked opinions of about ≈50 politicians, but our plot above restricts attention only to the six who contended for the 2017 Presidency. These six in total got 68+55+48+41+32+27=271% approval. There also were 5 other presidential candidates not included in the Paris Match poll (who, combined, got 4.04% of the official plurality votes). Some prominent noncandidates included in the Paris Match poll were (listed with their approval percentages)

Alain Juppe 60, Francois Bayrou 49, Segolene Royal 45, Francois Hollande 36, Nicolas Sarkozy 33.

Meanwhile the pollster IPSOS maintains approval ratings versus time data for many prominent French politicians online; we have listed many of those in the approval-summary table below, plus for the further non-candidates mentioned above IPSOS had

Juppe 50/40/10, Bayrou 38/48/14, Royal (unpolled), Hollande 23/73/4, Sarkozy 29/66/5.

One point of interest is that IPSOS's high approval rating for Melenchon and low disapproval for Macron during April and/or May 2017 were atypical versus other times (i.e. their approval or disapproval versus time plots have "spikes" when time≈April). Specifically:

This rapid time-variation makes it even more difficult to decide who "deserved" to win on election day, Melenchon or Macron – but it appears that in an overall time-averaged sense (disregarding the spikes near election day) France approved Macron more than Melenchon. And also Macron was more-approved than all his other rival candidates even without his spike.

Endorsements after the first round:

Fillon & Hamon: endorse Macron
Lassalle, Arthaud: endorse nobody
Melenchon, Cheminade: anti LePen
Dupont-Aignan: endorse Le Pen

Pairwise polls of the form "if the 2-candidate runoff came down to X versus Y, how would you vote?" were conducted by many pollsters. The following table summarizes the results of a large number of polls for all 6 possible pairings among the top 4 official finishers shortly before the first-round election day

XY
Macron 61-67%Le Pen 33-39%
Fillon 57.3Le Pen 42.7
Macron 65Fillon 35
Macron 60Melenchon 40
Melenchon 56Fillon 44
Melenchon 54-60Le Pen 40-46

and note the official second-round result (a month later) was Macron 66.1%, Le Pen 33.9%. Unfortunately I do not have any pairwise polls involving Nicolas Dupont-Aignan or Benoit Hamon, but French media claimed that Hamon's own privately commissioned polls showed that he would easily win the Hamon:LePen pair. In the official vote both Hamon and Melenchon suffered from a "vote split," but of course that problem would not have existed in a pairwise "Hamon vs X" or "Melenchon vs X" contest.

Those pairwise results indicate that the following honest-voter Condorcet order exists (at least among the official top 4):

Macron > Melenchon > Fillon > Le Pen    (and with Hamon presumably somewhere above Le Pen).

The team of academics consisted of

Jean-Francois Laslier, Directeur de Recherche en economie, CNRS, Paris School of Economics et ENS, Jean-Francois.Laslier@ens.fr
Antoinette Baujard, Maitre de conferences en economie HDR, Universite Jean Monnet, antoinette.baujard@univ-st-etienne.fr
Renaud Blanch, Maitre de conferences en informatique, Universite de Grenoble, Renaud.Blanch@imag.fr
Sylvain Bouveret, Maitre de conferences en informatique, Universite de Grenoble, sylvain.bouveret@imag.fr
Herrade Igersheim, Chargee de recherche en economie HDR, CNRS et Universite de Strasbourg, igersheim@unistra.fr
Annick Laruelle, Research Professor en economie, Ikerbasque, Universite du Pays-Basque, annick.laruelle@ehu.eus
Isabelle Lebon, Professeure des universites en economie, Universite de Caen Basse-Bormandie, isabelle.lebon@unicaen.fr

They studied both approval voting and score voting with score-levels either {-1, 0, +1}, {0,1,2}, or {0,1,2,3,4,5}. Their results as rank-orderings (and we have also included the official France-wide results and the Paris Match poll, although the latter was not conducted as a pseudo-election exit poll):

Voting systemRanking (top to bottom) output by voting system
Official Macron>LePen>Fillon>Melenchon>Hamon>DupAig>Lassalle>Poutou>Assel.>Arthaud>Chemnd
Approval Macron>Melenchon>Hamon>Fillon>LePen>DupAig>Poutou>Arthaud>Lassalle>Assel.>Chemnd
Paris Match Approval [or Score(0,1,99,100)?] Melenchon>Macron>Hamon>DupAig>LePen>Fillon (other 5 unpolled)
Score(-1,0,+1) Melenchon>Macron>Hamon>DupAig>Lassalle>Poutou>LePen>Fillon>Assel.>Chemnd>Arthaud
Score(0,1,2) Melenchon>Macron>Hamon>Fillon>DupAig>Poutou>LePen>Arthaud>Lassalle>Assel.>Chemnd
Score(0,1,2,3,4,5) Macron>Melenchon>Hamon>LePen>Fillon>DupAig>Poutou>Arthaud>Lassalle>Assel.>Chemnd

Their exit poll locations and the counts of those polled (versus the official counts of ballots at those same locations, which are greater because some voters opted not to participate in the "alternate voting systems" exit poll experiment) were:

Location Allevard-Les BainsCrollesStrasbourgHerouville/Saint-ClairGrenoble
Experiment 836261710717111069
Official 23085449187411802430

In all, they claimed that 6358 voters participated in the experiments, although the first row of the table above sums to only 6304. These participant-counts ranged from 36% to 60% of the total number of voters at those locations (on average 48% of them). All the non-official voting methods at all polling locations agreed with the Paris Match poll that the winner and second-place finisher should be {Macron, Melenchon} in some order, and that the 3rd-placer should be Hamon. However, who among {Macron,Melenchon} should win, depended upon the voting method and polling location.

On average over all 3894 approval-style ballots collected, 2.48 candidates were approved. The numbers of candidates approved were

#Candidates approved 01234567891011
On #Ballots 827511285108448913440166223

The 82+3 voters who approved either none or all of the 11 candidates effectively did not vote; this was 85/3894=2.2% of the voters.

Here are the approval percentages (corrected to apply, hopefully, to all France) by the academic team study as well as the Paris Match approval percentages, IPSOS favorable/unfavorable/don't know April 2017 approval "barometer" ratings, and official (first round) plurality vote percentages for comparison. (For both the academic study, IPSOS, and the Paris Match poll I estimate the ±1σ error bars should be roughly ±2%, e.g. for Macron regard 46.1 as "46.1±2.")

CandidateApproval% Paris MatchIPSOS "barometre" Official plurality
E.Macron46.15548/43/924.0
J-L.Melenchon39.46856/35/919.6
B.Hamon36.34839/50/116.36
F.Fillon30.12724/70/620.0
M.Le Pen27.53227/67/621.3
N.Dupont-Aignan21.14129/45/264.70
P.Poutou15.5 1.09
N.Arthaud8.7 0.64
J.Lassalle7.4 1.21
F.Asselineau6.1 0.92
J.Cheminade3.6 0.18

Conclusion

It is entirely possible that the candidate France most-wanted was J-L. Melenchon. If so, then the official system unfairly prevented his deserved victory. Clearly, Le Pen's "second place" official finish was far above the 4th-7th place finish she truly deserved. The underlying cause was that Melenchon and Hamon were hurt by "vote splitting" while Le Pen was a fringe candidate who suffered almost no vote splitting. (The official voting method is vulnerable to vote splitting, while approval and score voting are not.)

However, since the decision between Macron and Melenchon is unclear, it also is entirely possible that the official winner E.Macron really did "deserve" it.

Macron was a comparatively inexperienced newcomer to the French political scene. After the election, his approval rating sank rapidly during his first few months in office, reaching 40% in both late August 2017 and late March 2018. But Melenchon's approval also fell, to about 30%. And Macron's approval even at the reduced levels during his post-election year, still exceeded all his election-rivals. So at least as of the date of this writing (April 2018) it appears France still thinks it made the best choice among the available options.


France 2000+

Return to main page