I am trying to compute measures of all time spent “with” children for parents in the AHTUS – that is, any time during the day where a child under 18 was present – from all possible samples. I downloaded hierarchical data extracts and used the variable “child” (whether a child under 18 is present during activities) to identify activities where children were present. (I also manually coded activities where a child was not identified as present, but either secondary child care was performed or a child care activity was recorded). I limited my sample to only those with a child under 18 in the household (parents).
The total time for parents with children under age 18 is implausibly low for any year before the 1998 sample, and does not correspond with previous estimates. For example, parents report spending ~320 minutes per day with children in the 1998 sample, but only 65 minutes per day in the 1975 sample. In 1975, this estimate would mean parents only spent 7.6 hours per week with children. According to Bianchi et al. (2006), who use the same data, this should be anywhere from 21 hours/week (married fathers) to 47 hours/week (married mothers). Even without cross-checking this data, spending only around 8 hours a week with household children is not possible.
I have tried checking my own estimates (using the hierarchical data files, collapsed using Stata) against the IPUMS time use variable generator (selected all activities, if child under 18 is present, limited to households where a child under 18 years old lives). I get the same results, suggesting that it is not a coding error on my part.
I would very much appreciate any help on this issue, or who I can contact to discuss this! (I am hoping it is just some sort of misunderstanding on my part).
This sounds like it could be due to the known comparability challenges of the CHILD variable. Basically, this variable is created differently in some years. From 1965 through 1995 only a subset of activities were eligible to be coded as “with a child.” Beginning in 1998 is when the “with who” variables begin applying to any and all activities.
Thank you so much (belated!), Jeff, for the help with this. I have looked more into the available documentation and have an additional question.
Do you know if it possible to derive the presence of a household child using the variables SPPART and CLSFAM, for the 1975-76 sample? According to the documentation for the latter, “If the respondents said they were with a spouse/fiancé or with children
of the household, this variable [CLSFAM] is coded as 1.” As such, I am thinking that any instance of CLSFAM=1 where SPPART=0 would mean that it was a household child that was present.
If so – is there any available documentation on whether “household child” (as defined by CLSFAM) is limited to a certain age range? I looked at the original 1975-6 codebook and the SPSS files, but am not sure if IPUMS has any further information.
Thank you!
Based on the documentation for the SPPART and CLSFAM variables, for the 1975-76 samples, your intuition for identifying whether the activity took place with a child seems reasonable to me. This could be an alternative method for identifying the presence of children without using the CHILD variable. With that said, one limitation of this approach is there does not seem to be a carefully defined definition of the official age of “children of the household.”
My suggestion is to, when possible, use both the CHILD variable to directly identify the presence of a child and your indirect method using SPPART and CLSFAM. This will give you an idea of how sensitive your results are to these differing approaches.
Thank you so much, Jeff! You’ve been so helpful. I really appreciate it!