A simple two-stage data collection procedure to handle unit non-response in social surveys

The aim of this paper is to investigate the possibility of collecting information on the reproductive behaviour of mothers by means of a two-stages data collection procedure. At first stage a simple self-filled questionnaire is administered to a sample of students attending compulsory school; thus, each pupil is asked to give his mother a more complex questionnaire, bringing it back to school. Linking the two questionnaires, the non-response selection problems are solved using the same statistical techniques used to control attrition in panel surveys. Even if selffilled questionnaires can not be too complex, in some cases this data collection procedure can be recommended as (1) the proportion of unit non-responses in standard surveys can be very high, and the unit non-response selection is not easy to control; (2) it is a very low cost data collection procedure. After a general description of this data collection technique, an application is performed using data of the Urban Fertility Survey (UFS) conducted during 2001-02 in four Italian towns. As respondents are clearly selected (e.g., by citizenship) if data analysis is performed without correcting for non-response, results may be biased. In particular, we compare some regression models explaining fertility using data collected by the two questionnaires.