Age of Parent or Timing of Birth? Addressing the Challenge of Identifying and Interpreting Parental-Age Effects in Studies of Child Outcomes

Do children of older parents do better because their parents are older, or because they are born into “better times”? Recent debates on parental age and offspring outcomes suggest that this analytical distinction is harder to make than it may seem, and that existing analyses attempting this do not resolve the identification problem. The problem is that any serious attempt to address confounding must condition on parents’ birth year, which makes parental age at birth and child birth year perfectly collinear. I contribute to these debates in two ways. First, I formalize a general mechanism: under flexible, nonparametric specifications, the joint effect of two linearly dependent dimensions is mechanically reallocated in favor of the relatively more finely measured dimension, while their idiosyncratic effects remain unidentified. Using NLSY79 Child and Young Adult Survey (NLSY-CY), I illustrate this mechanism focusing on children’s educational attainment in early adulthood. Second, I recast parental-age analyses based on the Generations and Gender Survey as an age-period-cohort-type identification problem and adapt Fosse and Winship’s recently developed bounding approach to obtain partial but transparent identification of parental-age and child–birth-year effects on educational outcomes in Belgium, Germany, the Netherlands, and Sweden. Across a wide range of plausible assumptions, child-birth-year effects are non-trivial but cannot convincingly account for the key features of parenthood-age gradients, particularly the net advantages of later parenthood. Taken together, these findings add nuance to interpretations that period improvements largely explain delayed-parenthood advantages.