La Méthodologie des sondages électoraux de l'élection présidentielle française de 2007,chroniques d'un problème récurrent

Abstract

Election Poli Methodology and the 2007 French Presidential Elections - Chronicle of a Recurring Problem: The 2002 presidential élection polls will be remembered for a long time because their error in predictions had dramatic consequences. What influence did those elections have on French opinion poli research and was it able to revise its methods? Was it able to improve its voting estimates? Research has shown that French polis are now more able to distinguish between voting for the left and the right than in 2002. However, French polls have not improved their capability to estimate intentions to vote for major candidates. The discrepancies between means for twelve polis published during the last week before voting vary from 1.8 points for Royal to 3.2 points for Sarkozy and 3.8 points for Le Pen. These results show that the estimation error for Le Pen is the same in 2007 as in 2002, but in the opposite direction. As in 2002, samples were biased concerning level of education - strong under-representation of the less educated - as well as for voting profile - strong non-declaration of voting for Le Pen. One can thus conclude that the problem of estimating the extreme right vote remains in its entirety and that polling institute methods have not changed greatly. Cooperation between the institutes and researchers should permit further explanations and possible solutions.