I attached data for mothers and fathers to a CPS sample of children 0-19. For children with “mother” and “father” age variables with known values, the age gap between children and their mothers or fathers is sometimes less than ten years or more than 50 years. Can we trust the algorithm that attached these parental data?
The variables MOMRULE and POPRULE (MOM2RULE and POP2RULE for same-sex parents) report the certainty around parental assignments using the IPUMS algorithm. This is a useful tool for assessing if the parental assignments seem plausible/appropriate for your research applications. You could also check the IPUMS-generated family interrelationship variables against the self-reported variables in the original CPS data (PELNPAR1, PELNPAR2, PELNMOM, PLENDAD)–recognizing that the self-reported variables report the line number (LINENO) of the parent while IPUMS variables report the person number (PERNUM) and the two may not always be the same.
For the age differences you are describing, I can immediately think of two situations that would result in these types of age differences you described. The first is direct links with unambiguous relationship values (e.g., a “child” and “householder” are linked regardless of age differences as the relationship is clear and direct). The second is when there is a large age gap between spouses. If a person has an assigned spouse (via SPLOC) and is also identified as a probable parent to someone else in the household, that person’s spouse will automatically be assigned as the second parent. If one parent has a “reasonable” age difference with the child but the age gap between that parent and their spouse is quite large, you would see such an age difference.