The recast population census introduced in 2004 is a rotating sample operation which annualises data collection, expenditure and the publication of results, while ensuring improved efficiency for the different players. The main feature is that only a fraction of the population of households in municipalities of 10,000 inhabitants or more are surveyed: thus the sampling is one more factor affecting the quality of the traditional exhaustive census. This decision raised many questions about the added variability introduced for estimates based on these new census data. In particular, local officials were worried about the impact of this new feature on official population statistics, on which the amount of central government funding awarded to municipalities is based. The aim of this article is to give an overview of studies that have quantified the impact of sampling on the accuracy of the census results, and to describe the situation as it stands today. There have been two phases of studies. First, precise calculations were carried out via simulations based on data from the exhaustive 1999 census. Next, these initial results were refined by performing calculations on the first available data from a complete cycle of the renovated census. These methodological studies show that the scale of this inaccuracy is negligible and hence the move to data collection by sampling has had no deleterious effect on the quality of the census results.