Time with children inconsistency


I use the ATUS and AHTUS data regarding time spent with children over the different surveys (both total time spent, defined as having a child present). When examining the means of these variables (only for households for which this variable is greater than 0), there seems to be a spike from the 1998 sample onwards. In the 1995 survey, the mean was about 106 minutes, while for the 1998 survey it was 311 minutes (with the means remaining consistent in the 300+ values ever since).

As far as I know, there is no change in the variable description since 1998, and I’d like to ask if this inconsistency could be attributed to a different measuring, sampling or calculation method.

Thank you very much.

It sounds like you are interested in potential comparability issues for the variable CHILD (Child under 18 present during activity) in AHTUS, please correct me if I am wrong. Due to the fact that the AHTUS is a harmonized collection of historical time diary surveys dating back to 1930, there are comparability issues for many variables that should be considered in any analysis that considers changes over time. You may want to take a closer look at the Sample Descriptions for each survey to determine how changes to sample design could impact your analysis. In the case of CHILD, there are differences in how CHILD = 1 (“yes”) is coded from the 1994-1995 National Time-Diary Survey to the 1998 Family Interaction, Social Capital, and Trends in Time Use Survey. In 1995, CHILD was coded as 1 during activities related to baby care, child care, helping/teaching kids, indoor/outdoor playing, medical care for child, and other general care for children. Conversely, in 1998, CHILD was coded as 1 if the respondent reported the presence of a child in the “who with” variables (for more information, see the comparability tab for CHILD).

Thank you very much for your response.
I also try to focus on specific activities, which also present a similar pattern of a sharp jump between the early 90s to the late '90s (and afterward, including the newer ATUS data). This leads me to another question - is it possible to differentiate between primary and secondary activities, thus possibly overcoming the different coding?
Thank you.

You can differentiate between primary and secondary activities in a hierarchical extract using the variables MAIN and SEC. However, there is very little detail in SEC for the 1995 sample, so I don’t think this will help you very much.