Thursday 27 April 2017
“I cannot prevent the French from being French.” - Charles de Gaulle
Are the French elections a one-, two-, three-, … candidate race?
By Jocelyn Evans, Gilles Ivaldi
03 February 2012 | Polls & Forecasts | 1044 words
If you (foolishly) rely solely upon polls to understand the state of play in an election, which polling house you choose will give you a biased view of that position. ‘House effects’ determined by sampling frames, response rates, ex-post adjustments to correct idiosyncracies of parties or candidates revealed in previous polls, and even the question asked of respondents will all give different pictures of where each competitor stands in relation to their rivals. Looking at the different French polls at the moment reveals more than just minor discrepancies: at the extreme, they seem to be covering different elections.

Compare for instance the 31 January rolling poll by IFOP- Paris-Match, BVA cross-sectional poll carried out on 30-31 January and the OpinionWay poll of 26 January.

  BVA IFOP Opinion Way
François Hollande 34 31 27.5
Nicolas Sarkozy 25 23.5 24
Marine Le Pen 15 20 17
François Bayrou 12 12 14
Jean-Luc Mélenchon 8 7.5 8
Eva Joly 3 3 3
Effective number of candidates 4.5 4.7 5.3
Mainstream* 74 69.5 68.5

* Hollande + Sarkozy + Bayrou + Joly

Ostensibly, at around 90 days from the election, there are three very different competitions in evidence. For both IFOP and BVA, the Socialist candidate Hollande has a significant lead over the presidential incumbent. For OpinionWay, the margins are much tighter, and well within the margin of error – in the first round, at least, far from a given who will win.

Secondly, the Extreme Right candidate Marine Le Pen has been presented as a clear danger to the very same incumbent president in the now clichéd ‘mirror image of 21 April (2002)’, where she moves into second place and the second round, against Hollande. For IFOP this is a very real possibility with only a 3.5 point difference – and a couple of days previously in the same daily rolling poll, a 1.5 point difference, with Le Pen on 21 percent and Sarkozy 22.5. (Although, in passing it should be noted that one of IFOP’s faults is the day-by-day variation its polls produce, suggesting all manner of dynamics many of which are simply noise.) For BVA, the likelihood of a Left-Extreme Right run-off is even more distant than a Sarkozy first-round victory. What is more at stake for Le Pen in this poll, and in OpinionWay’s, is a very much identical image of another 22 April (2007), where the Front national candidate is knocked back into fourth place by the Centrist François Bayrou. For IFOP, Bayrou has a mountain to climb to reach third place.Finally, what of the radical Left Jean-Luc Mélenchon? He clearly occupies fifth place in all three polls, but for two is within margin of error of fourth place. For these, a repeat of 2002 where the Extreme Left reached double figures is not impossible.

If we focus on the relative strength of competitors and where the narrowest margins are to be found, BVA currently sees the fiercest competition for bronze, with gold and silver decided; IFOP sees silver under contest, but again regards gold as a given; OpinionWay sees the clustering at the top of the table, focusing on the gold medal. Another, less Olympian way of looking at the same data is to calculate the fragmentation of the presidential race. This looks at the ‘effective’ number of candidates, which is the adjusted number of candidates according to their relative electoral strength (see Note below). OpinionWay predicts a more fragmented race, given the clustering of candidate scores whereas both IFOP and BVA would anticipate an election very similar to the 2007 presidentials, where the effective number of candidates settled at 4.7. However, even OpinionWay does not come anywhere near the fragmentation of the 2002 Presidential race, which reached 8.6 candidates.

Another parallel with 2007 can be found in the mainstream candidates achieving about 7 in 10 votes in the first round. Within this candidate bloc from ‘parties of government’, it is clear from all polls that Hollande has replaced Sarkozy in his former functional role of ‘candidate for change’ – hence, Sarkozy’s latest and apparently desperate attempt by an incumbent to revert to the ‘dark horse’ challenger role. Secondly, political discontent and the economic crisis seem to be taking voters away from the ‘soft’ anti-system line symbolised by centrist candidate Bayrou in 2007 and rebalancing towards the ‘hard core’ populist campaign by the FN.

In terms of both polling scores and fragmentation, further changes are obviously to be anticipated. These scores represent the journey, rather than the destination. The direction of polling depends on the campaign and the electorates reaction against it. Fragmentation is very unlikely to increase; rather, following Chevènement’s example, a number of marginal candidates will in all likelihood abandon the race in the next few weeks.

The campaign then has two possible routes: either the competition will focus on the Hollande/Sarkozy duel, thereby continuing on the bipartisan track of 2007, or, conversely, Bayrou and /or Le Pen manage to gain more political traction and widen their electoral support before April, in which case the race will line up closer to the array of four competitors of roughly equal size, characteristic of the ‘quadrille bipolaire’ of the late 1970s. In either case, the fifth candidacy of Mélenchon simply colours in the left flank, but is very unlikely to alter fundamentally the competition. In an interesting mirror image of the 1970s, Le Pen’s shifting economic policies to the left would confirm further that the FN intends to replace the old communist party in its ‘fonction tribunitienne’ embodying a ‘negative’ counter-power within the party system. As for Bayrou, only time will tell whether he is to return to his home ground at the centre-right of the political spectrum.

The hefty differences by polling house in who is tailing whom belie, then, a relatively stable picture of a Left-Right contest, either between two candidates or two, two-candidate blocs. Neither of these resembles 2002 more than 2007 in the shape of the line-up. On this evidence, then, we would not currently be inclined to support the notion of a disruptive third-candidate presence on 22 April. A two-candidate race; a four-candidate race; but in no sense, a three-candidate race.


The effective number of parties was introduced as an index by Mikael Laakso and Rein Taagepera in their 1979 classic, “Effective Number of Parties: A Measure with Application to West Europe” in Comparative Political Studies (12:3-27). It is formally defined as the inverse of the sum of squared individual party proportions.


Welcome to '500Signatures', for analysis and commentary on French politics and elections

This blog is produced by Jocelyn Evans (University of Leeds) and Gilles Ivaldi (University of Nice)

Politicizing terror: terrorism and the 2017 presidential race
Hollande’s calculation behind the French socialist presidential primary
Is the French 2017 presidential battle already over?
Estimating Marine Le Pen’s 2017 presidential vote share
Walking a fine line? Hollande and the French Left
All roads lead to Rome: French parties on the way to the 2017 presidentials
Ils ne passeront pas – the stemming of the FN tide in the regional run-offs
The FN on the threshold of regional government
Regional elections and the anti-Muslim backlash
Politics in a time of war?
A right-wing landslide but no far-right tsunami: the departmental election run-off
The Front national is not France’s first party
What to expect in next month’s French departmental elections
Departmentals 2015: the new French elections no-one seems to care about
Beta-testing social-liberalism 2.0
France’s new earthquake election? The FN in the European elections

posts have been published
since 10 January 2012

Show all posts

Jocelyn Evans [@JocelynAJEvans] is Professor of Politics at the University of Leeds

Gilles Ivaldi is a CNRS researcher in political science based at the University of Nice



- Forecasting the FN vote in Second-Order elections (updated 12 May 2014)

- Forecasting the FN vote in Second-Order elections (Jan. 2014)

- Polling scores by polling type (CATI v CAWI) (updated 20 April 2012)

- Estimating Marine Le Pen's vote in the 2012 presidentials: an experiment (November 2011)

- Data for the 2011 expert forecast survey (in CSV file)



Last modified on Monday 25 April 2016
Copyright Gilles Ivaldi - @2012-2014