Statistical Challenges in Identifying Effects of Age and Age Differences

Christian Dudel , Max Planck Institute for Demographic Research
D. Susie Lee, Max Planck Institute for Demographic Research
Angela Carollo, Max Planck Institute for Demographic Research
Nis Brix, Department of Public Health, Aarhus University
Maria C. Magnus, Norwegian Institute of Public Health
Mikko Myrskylä, Max Planck Institute for Demographic Research

The age difference between two persons is often of interest as an exposure in epidemiologic research. The association of the age difference between mother and father and child health outcomes is an example. In this paper, we show that there are three issues when using age differences as exposure. First, it is not possible to simultaneously control for the underlying age of both persons and the age difference without introducing arbitrary and untestable assumptions; instead, only two of these three variables can be accounted for at the same time. Second, we show formally that the age difference does not capture any interactions between the underlying ages; this might seem counterintuitive, given that the age difference is calculated using the age of both persons. Third, we argue that age differences might lack any deeper substantive meaning, at least in some applications. We illustrate these points using U.S. birth register data on more than 3 million births and modelling the influence of the maternal age, the paternal age, and the parental age difference on the risk of low birth weight; using this example, we provide recommendations on how to properly model and represent the interaction between age variables.

See paper

 Presented in Session 112. Survey Mode Effects and Measurement Challenges in Demographic Research