The Contribution of Residential Mobility to Sample Loss in a Birth Cohort Study: Evidence from the First Two Waves of the UK Millennium Cohort Study
Ian Plewis, Sosthenes C. Ketende, Heather Joshi, Gareth Hughes,
Longitudinal surveys gain measurement occasions but lose cases over time. Knowing more about reasons for sample loss improves the chances of being able adequately to adjust for it. This article shows that residential mobility is an important predictor of sample loss over the first two waves of the most recent UK birth cohort study, the Millennium Cohort Study. The measure of residential mobility used takes account of moves after wave one using the information available from the administrative side of the survey operation. We find that the residentially mobile are more likely to be nonrespondents even after controlling for a range of background variables. Mobile households are, however, somewhat less likely to refuse than the nonmobile and mobility is unrelated to noncontact, provided the sample cohort member has been located. The implications of the results are discussed in terms of the assumption of data being “missing at random” in longitudinal surveys.
Longitudinal surveys, attrition, nonresponse, moving home, logistic regression, multilevel models, UK Millennium Cohort