Problems in the family: controlling for age, period or cohort in sibling comparison designs
Ben Wilson1, Maarten Bijlsma2, Siddartha Aradhya3, Alice Goisis4
1Stockholm University, 2Max Planck Institute for Demographic Research, 3Lund University, 4University College London

The sibling comparison design is increasingly used across the social sciences a means of controlling for confounding. Accompanying this trend, is a growing body of research that highlights methodological challenges when using sibling comparison designs. We add to this literature by showing that –under certain circumstances– it is possible to obtain biased estimates when controlling for age, period or cohort in a sibling comparison design. We demonstrate these circumstances using a simulation study, which allows us to compare different estimators and their relative ability to obtain unbiased estimates. Our simulation illustrates that this issue is not the same as the classical issue of APC collinearity, but instead is a result of making within-family comparisons. It appears that one of the most promising means of avoiding bias is to control for age, period or cohort using a group that are discordant on these variables but concordant on the exposure.