Hi IPUMS time use community,
Using an extract of 2003-2023 ATUS data that includes ATUS respondents and their household members I noticed discrepancies between information derived from the AGE and a lot of the “constructed household composition variables” and “constructed child variables.” For example, there are 1643 households with a case aged under 18 within households coded as hh_numkids=0. That is, hh_numkids codes quite a few households as having no kids under 18, while according to the AGE variable they do have kids under 18 in them.
Additionally, comparing AGEYCHILD with the actual age of the youngest child in each household using AGE within households with kids under 18 shows that there are quite a few households (261 to be exact) in which the age of the youngest is lower than the age reported in AGEYCHILD (it is as if AGEYCHILD just “missed” the youngest child). The same is true for KIDUND1, KID1TO2 and so on.
Can it be that ATUS includes rows that are not considered household members in the constructed variables? What am I missing here?
Any answer will be much appreciated and thank you for your time!
All the best,
Amit.
The IPUMS ATUS variable HH_CHILD reports whether there are any children under the age of 18 in the household. The variable HH_NUMKIDS reports the number of children under the age of 18 in the household.
Children that belong to the ATUS respondent, but who are not part of the household, also have person records in the household roster. Therefore, there can be disagreement between the number of children according to HH_CHILD and HH_NUMKIDS versus according to the household roster. You can identify these nonhousehold children of the ATUS respondent using the RELATE variable. If you exclude nonhousehold children (exclude everyone with RELATE=40), you will see that there is agreement between the household roster and the variables HH_CHILD and HH_NUMKIDS. This also resolves discrepancy between the household roster and AGEYCHILD.
The variables KIDUND1, KID1TO2, etc. are different, however. These variables report the number of own children of specific ages in the household. Only ATUS respondents (those with RELATE=1) have responses for these variables (see, for example, the universe of KIDUND1), so they only list the number of children of each age that belong to the ATUS respondent. Other household children—those who are not the children of the respondent—will not be counted in these variables. There may be many instances where there are children that do not belong to the ATUS respondent in the household.
Thank you very much for your reply Isabel! I now understand that own non-household children of the ATUS respondent are included in the household roster. In that case, I would suggest adding a note in the description of HH_CHILD, HHNUMKIDS and AGEYCHILD explaining that non-household children are not counted in these variables but are included in the data. This is extra confusing since HH_SIZE, for example, does include these children in the calculation (as far as I can tell) but has the same description of “people living in the respondent’s household.”
A short follow up question: Do you happen to know the BLS definition for “non-household children” beyond the not very informative definition of: “a respondent’s child under the age of 18 who does not live with the respondent”? (what if they sleep in the household some of the days of the week? Or on weekends?)
Thank you very much for your time, effort and help!
All the best,
Amit.
I will pass along to the IPUMS ATUS team your suggestion to note in the metadata that these variables include non-household children of the ATUS respondent. I agree that would be helpful, since it is not obvious.
I have not been able to find any documentation that provides more detail than the sentence you quoted about how non-household children are defined for the ATUS. I agree the definition is vague, and in that way is left to the interpretation of the survey respondent. In the ATUS questionnaire, there is a question that asks, “Do you have any children under 18 who do not live with you?” This question is used to determine the addition of any non-household children to the ATUS household roster. It is not more descriptive than the definition of non-household children you already have. I would interpret this as meaning that an individual parent’s definition of whether their child lives with them is used to determine who is a household child versus a non-household child.
Thank you very much for your replies! They have been very helpful!
All the best,
Amit.