I attached ages of coresident fathers and mothers in the final step of submitting a request for ACS data. Even for the “purest” allocation (MOMRULE==11) I find that the difference between ages of mothers and ages of children decreases as child’s age increases. There are cases of 15-year-old children with coresident “mothers” of age 15! Clearly there is a flaw in the allocation of parental records to children in a household. Thoughts?
There are a few things that could be causing this type of irregularity. First of all is that step parents can be linked as parents, and so would not need to necessarily be older than their step child. Another issue is that RELATE codes are sometimes allocated by the Census Bureau, so it is possible that some of the links were not accurate. The Census Bureau also modifies the public use microdata in certain ways including swapping values of variables between cases to preserve confidentiality, so this can introduce illogical patterns. Fortunately the number of cases like this is quite small. If you’d like to look into this more, I suggest using the quality control flags (QAGE and QRELATE) to identify allocated cases of AGE and RELATE. Please note that these allocations are done by the Census Bureau before releasing the public use files - IPUMS does not modify these variables.
Thank you for your reply! This is very interesting, but still leaves the “parent” matching somewhat suspect. I’ll make do somehow!
Using CPS 2000 data, I have 29 records for children under age 15 who report being child of the householder but have no linked records present for mother or father. Any suggestions about how to classify such cases where a child of someone in the household has neither father nor mother records linked?
I downloaded an extract containing all of the 2000 CPS samples, and I did not find any such cases (where RELATE=301 and MOMLOC=0 and POPLOC=0). Can you give some particular cases (identified by year/month/serial/pernum) where you are seeing this, or clarify exactly how you are identifying these cases?
Thank you for your reply, Matthew. From what you say in your 5/17 email, the cases where I have no mother or father present for children of household heads must be some of the few cases that I have NOT coded as mothers or fathers because their ages compared to the child’s age make such parental relations impossible. (For example, a “mother” with the same age as her “son” or “daughter.”) I will check to see if this is the source of my uncoded cases, and let you know. I continue to think it is counterproductive to code parent/child links between people of the same age, and hope that someday such refinements make it into the public use microdata files so that we can trust we reallly have “parents” and “children” to study.