The aim of this paper is to discuss and try to address the unit non-response selection problem in the case of a population of interest composed of mothers. A two-stage procedure is used. In the first stage, a short self-compiled questionnaire is administered to a sample of school-children aged 13. They are then asked to hand to their mothers a more complex questionnaire to be filled in an brought back to school. By linking the two questionnaires, we manage to solve non-response selection problems by means of the same statistical techniques used to control attrition in panel surveys. This type of data collection procedure is recommended for the following reasons : (i) the proportion of unit non-response in standard surveys is sometimes very high, and this can lead to a non-response selection bias which is not easy to control ; (ii) the cost of this procedure is very low. This technique was applied to the Urban Fertility Survey (UFS) conducted in 2001-02 in four Italian cities.