I have read some of the “Not In Universe” Descriptions for variables in the ATUS. I understand that some terms will not apply due to being blank or not applying, but why would Month of the survey be out of universe? Other variables are included, so something was filled out in the survey. What would have to happen for us to not still have the month of their reply unless they left everything blank?
I suppose a follow up or related question is about the construction of SERIAL. There are a handful of observations in 2018 and 2017 that have the same HRHHID_CPS8 but have different SERIAL.
If SERIAL is supposed to be a unique household indicator, and PERNUM is supposed to identify the person in the household, then how would two observations have the same personal identifier but have different household serial identifiers?
All variables have a “Universe” tab, which specifies who is included in a given variable. The universe for MONTH is “ATUS respondents.” However, IPUMS ATUS allows you to include household members (i.e., who reside with the respondent) as well as non-respondents (i.e., those selected as the respondent but who did not complete the ATUS) in your extract. If you look at the codes tab for MONTH, you will see that there are not any NIU cases if you restrict your sample members to respondents only.
The CPS household ID has two parts: HRHHID_CPS8 and HRHHID2_CPS8. I suspect using HRHHID_CPS8 and HRHHID2_CPS8 together should address this issue.