|Much has been made in recent days of Nicolas Sarkozy overtaking François Hollande for the first time in polling for the first round of the French presidentials. This symbolic croisement may or may not reflect a change in Sarkozy’s fortunes and the first step towards an historic electoral turnaround. It certainly reflects a trend to date found only in the IFOP polls, well within the margin of error and not replicated in the second round polls, which still all give Hollande a clear if reduced victory. This campaign key moment has overshadowed a potentially more interesting – and troubling – polling phenomenon overlooked by most of the press. |
On 14 March, the French press reported that an IFOP poll of les primo-votants – first-time voters – put François Hollande way out in front, with 31% of vote intentions, followed by Marine Le Pen (23%) and Nicolas Sarkozy (21%). The same day, a CSA poll carried out amongst those in the 18-30 age category concluded that Hollande was out of step with the young electorate, managing only 18% of this group, against 26% for Le Pen and 25% for Sarkozy. On Monday evening, amongst 18-24 year olds, an ostensibly earlier IFOP poll had 27% for Hollande, 26% for Sarkozy and 16% for Le Pen.
||IFOP, 14 March
(18-22 years old)
|CSA, 12-13 March
(18-30 years old)
|IFOP, 11-12 March
(18-24 years old)
First, a note on the respective methods, as reported by the polling institutes. The CSA poll was a national sample of 861 over-18s interviewed by telephone; the earlier IFOP poll a similar national sample, but about twice the size – 1638 respondents – carried out online. The IFOP poll reported on 14 March was from a restricted sample of 805 18-22 year olds, carried out both online and by telephone. First point of interest: according to the fiche technique, this had been carried out between 22 and 28 February – some three weeks prior to the other two polls. Reporting this lead of Hollande’s as ‘current’ was thus somewhat misleading on the part of the French press, none of which gave the dates of the survey.
Even so, a plunge of 13% in even three weeks is a remarkable drop for any candidate, especially given the interim poll by IFOP still placed Hollande at 27%, a mere two days before the CSA raz-de-marée. A possible conjecture would be that the differing age bands are responsible. IFOP’s first survey (the one reported second) looks only at those voters who were ineligible in the 2007 election. The second survey (reported first) extends this to 24 year olds, and the CSA sample includes a further six years – roughly, those voters who were in the 18-24 band in the 2007 election. On that basis, then, if samples are representative and we accept the IFOP figure of 27% as accurate within the bounds of expected error, the 25-30 age group needs to account by itself for most of a 9% swing against Hollande.
This in itself is staggering. According to the Panel Electoral Français , in 2007 the youngest group (18-24) split 22% for Nicolas Sarkozy, 31% for Ségolène Royal. A group which reflected a higher level of support for a Left-wing candidate five years before has now moved sufficiently away from that candidate to offset new voters to a point well below that of a Right-wing candidate noted at the time for his lack of support amongst the young. Looking at the 2008 French census data and assuming a stable proportion of 1 to 1 for 25-30 years olds and 18-24 year olds, if the 27% figure published by IFOP on 12 March is accurate, then the 25-30 year old group can only support Hollande at around 9% - between three and four times less than support for the less popular Royal in 2007. This is way below even what a minor candidate like François Bayrou polled in 2002, let alone 2007, amongst the youngest age group.
Admittedly, the category banding introduces a year’s overlap on our calculations, margins of error will mean that the contrast is not as stark as this – a point which newspapers would do well to take into account when heralding the historic croisement – and given the much earlier date of the primo-votant IFOP poll, Hollande’s polling scores will certainly have fallen among some if not all age groups. Furthermore, as Anne Muxel has pointed out, the youngest voters in France apparently represent a new group of independent, non-partisan voters who, even by the standards of unstable first-time voters, are protean in their choice and their stability.
Nevertheless, for polls proclaiming to be representative, the sheer size of the disparities on display beggars belief, whatever the group dynamics of younger voters. Can a five-year gap really produce such a change in support for the mainstream Left, particularly in a period which has seen high levels of support for that candidate? Moreover, is it reasonable to believe an 18-point differential between two adjacent age-groups? However, as we have noted before, until French polling institutes are more open about their sampling and sample corrections, it is impossible to work out what is a true dynamic within the population and what is merely artefact. Fundamentally, is IFOP too high or CSA too low? If polling institutes wish their data to be taken seriously, they must be willing to acknowledge and try to explain such disparities. In an interview with leJDD.fr after their infamous poll hypothesising Marine Le Pen’s absence from the first round, Frédéric Dabi, IFOP deputy-director, explained that “nous ne sommes pas là pour rendre service ni pour créer l’information, mais pour refléter un état de l’opinion publique”. “We are not here to provide a service or to create information. We are here to reflect the state of public opinion.” The reflections currently on offer are reminiscent of a hall of mirrors.