Within teams responsible for national election surveys there is much debate about design, especially interview modes. Most often, gaps between the theoretically ideal and the empirically manageable fieldwork are due to costs or sampling methods requirements. For instance, in France, random sampling is quite a challenge - with the remarkable exceptions of the 2000 Democracy Survey and the 2007 CSES project, both implemented with phone interviews. Yet survey design has crucial consequences not only on questions format but also on analyses which are (im)possible to perform afterwards. Within this framework, our paper will focus on a specific aspect of survey design: the geographical spread of respondents and the subsequent (in)ability to fully grasp the local and social contexts influencing their voting decisions. However, thanks to CARTELEC, an on-going project we manage with geographers, innovations in surveying French voters could be envisioned. With a new strategy for contacting respondents, a future electoral survey could seek a representative sample structured so as to enable multilevel analyses, possibly even randomly selected. We will take as a starting point the link between the localized / national spread of respondents and the models of vote choice built upon such surveys since the pioneer work by Lazarsfeld and in the ANES series. We will then review past French national electoral studies where “atomistic” designs - purposively avoiding clusters and correlation effects - could not adequately tackle the impact of local factors, even though improvements have been sought lately. Last, we will explain how our cartographic project CARTELEC, in which aggregate results are combined with socio-economic variables at the refined level of polling places, could help building a national representative sample stratified differently. This would, for instance, imply “itineraries” to contact respondents for F2F interviews. Would the added value in statistical power be worth the costs of implementing such a fieldwork for the 2012 French election study?
