IPUMS_USA: is it possible to know if a person cohabits?

The marital status variable has six categories, which does not include cohabitation, and the relation to household head variable has one category lumping partner, friend and vistor all together as one. In this case, partner does not necessarily mean cohabitate partner, right? I’m wondering if there is way we can know a person cohabits from this dataset? Is there any other variable that could be helpful to distinguish cohabitation?

I appreciate your help!

It is possible to identify cohabiting relationships that include the head of household. You can identify such relationships by utilizing the detailed version of the RELATE variable. A person with a RELATED value of “1114” is the unmarried partner of the head of household, while a person with a RELATED value of “0101” is the head of household. Unfortunately, there is no way to confidently identify cohabiting relationships that do not include the head of household.

Note that you will need to click the “Detailed codes” button on the Codes page in order to see the labels for the detailed version of RELATE.

Hope this helps.

Not an answer, but a follow-up question. Thank you, Tim! It does help! I check the availabilty of unmarried partner, and it’s only avaiable for sample since 1990, and it has to be 1990_5%. I’m trying to do a comparison study from 1980-2010, and for 1990, I’m using 1990_1% sample. I guess I probably cannot be able to know cohabitation for 1980 and 1990 sample, right?

The “Unmarried Partner” code is available in all 1990 samples; however, it is not comparable with the 2000 samples. In 1990, same-sex spouse reports were recoded in one of two ways: one of the respondents’ sexes was recoded or the “spouse’s” relationship was recoded into another relationship. In 2000, this policy was changed: all same-sex spouse responses were recoded as “Unmarried Partner” responses. Due to this difference in coding same-sex relationships, the “Unmarried Partner” category is not comparable between 1990 and 2000. See this page for more comparability issues with the RELATE variable.

Unfortunately, for 1980 samples it is not possible to accurately measure cohabiting relationships, since partner and roommate are combined into a single category.