ABSTRACT
The advent of probabilistic sampling methods has been commonly hailed as a revolution in the work carried by government statistical offices. The idea of sampling, which runs counter to a century of efforts devoted to the improvement of statistical coverage by administrative control and cross-checking, could not be widely trusted before the community of statisticians had reached an agreement on strictly defined standards. The author examines the limits imposed upon the imagination of statisticians by the nation-wide character of the census and the constraints, it induces to their ideas about representativeness. With regard to the social division of statistical labour, the introduction of sampling methods implied an intellectual approchment between academic mathematical statistics and government statistics. Moving to large-scale survey sampling means setting up methodology and research departments within the statistical office and staffing these with highly trained scientific personnel, as well as reorganizing the whole routine of economic and social data collection.
