Abstract:

The 1986 french "Elections legislatives" have pointed out that the figures out coming from opinion polls are highly dependant from the statistical techniques involved in the estimation process. Such influence can be observed at three levels : - Questionary design and fieldwork protocol; - Reweighing of the sample to correct bias on past political events recovered from memory; - Simulation models used to forecast the outcome of the elections. The object of this paper is to account for what IPSOS and G-LINE have learned on such matters during the joint research they undertook under the particular circumstances of the latter french election.