Choosing who to follow: The long-run impact of following rules on the sample size and composition of household panel surveys

Abstract

Following rules for a household panel survey determine who is followed and interviewed when there are changes to the household structure over time. To understand the long-run impact that various following rules have on the size and composition of the sample, a household panel sample is simulated such that every household split-off is developed over a 40-year period. This long time period allows sufficient time for children to be born into the simulated sample and subsequently leave home and have their own children. Following every split-off allows for the application of the widest following rules along with a range of narrower following rules. Holding everything else constant (such as non-response rates, propensities for household formation and dissolution, fertility rates, death rates, etc.), the widest possible following rules result in a sample that is much larger and has quite a different composition to the samples with narrower following rules. Even within the narrower following rules, some small differences in sample size and composition appear by wave 16 and these become more apparent by wave 41.