CitationUdry, J. Richard & Chantala, Kim (2003). Missing School Dropouts in Surveys Does Not Bias Risk Estimates. Social Science Research, 32(2), 294-311.
AbstractObjectives. Our purpose is to advise those who propose to use school-based sampling frames in studying risk attributes of the adolescent population. Compared to using area sampling frames, will their estimates be biased by the fact that they do not have data from adolescents who are not enrolled in or present at school, primarily dropouts and graduates?
Methods. We used a national probability sample of school enrollment rosters as a sampling frame but administered questionnaires in respondents’ homes. In this way we obtained data not only from those attending school but also from those who had dropped out or graduated between the time rosters were compiled and the time questionnaires were administered. We compare estimates of attributes with and without dropouts and graduates.
Results. Estimates based on omitting those who had dropped out or graduated within the year before interview are only trivially different from estimates that include them, even though dropouts and graduates often differ significantly from those still attending school.
Conclusions. Although dropouts and graduates are missing, school administration of interviews does not significantly bias population-wide estimates. This is because dropout rates at the national level are currently quite low. Although they differ significantly from the enrolled, the absence of dropouts does not bias estimates for the total population.
Reference TypeJournal Article
Journal TitleSocial Science Research
Author(s)Udry, J. Richard