Conferences

Quantitative research requires probability sampling: that is a commonly accepted principle. Must we, however, always set aside the idea of a survey without sampling? This paper presents a survey that seeks to address this challenge.

ELVIRE is a survey conducted on 2006-2008 by the French National Demographic Institute, in order to investigate the languages used by the researchers working in the French public research Institutes and Universities. The survey has two parts, each with its own self-administered web questionnaire. The first part (called Unit) aimed at surveying exhaustively the 4000 heads of the public research units; the second part (called Individual) targeted all researchers and PhD students working in these units.

The starting point of the sampling was a non-updated and incomplete database of the research units, including the name of their heads and some other information. Therefore the crucial task was update the units’ database, necessary for the first part of the survey; and to find the lists of their members, necessary for the sampling of the second part. The first point was successfully completed and 54% of the heads filled out the questionnaire available on-line between December 2007 and March 2008. The second part, aimed at establishing an exhaustive list or a representative sample of the target population of the second survey, turned out to be impossible.

The paper will describe the strategies implemented to obtain a database or a representative sample of the research units’ members: requests to Universities and research Institutes to send us lists; requests to units’ heads to do so or to indicate a small number of persons according to objective criteria, etc.

The different efforts were not successful but they provided us with one third of the member lists even if not always completed and statistically representative. Thus, the Individual part of ELVIRE was conducted as an open, voluntarily jointed, survey based on a positive starting point of two databases: the list of units (with fitting information) and the members’ lists already collected. Finally, almost 9000 questionnaires were filled out between November and December 2008.

The paper will also present the setting up of this open survey in order to maximize its diffusion; contacting people as different as possible; estimating the distance between our respondents and a representative sample. So, a special effort was made to diffuse the information to all the research units using the heads’ database. At the same time, we informed all the researchers listed in the second database (almost 35 000 persons) and diffused the information across the Institutes, universities, professional associations, scientific libraries etc. Some questions of the questionnaire were designed to measure the profile and the degree of interest of the respondents. Finally, it will try to compare the respondents’ profile with some already known external data on the composition of the research community.