OFFPOV - Conflict on poverty status within household in ASEC 2021

Hi there,

I am interested in studying how household composition (in particular % of household employed or in labourforce) effects household poverty rates.

In the ASEC 2021 data I have encountered an issue where there are households identified by their serial number where individuals within the household have different poverty statuses. From reading FTYPE, my understanding is that these are 2 separate family units within the household. Is my understanding correct?

Following on from this, how do I uniquely identify each family household that OFFPOV is applied to for the purposes of my analysis and the number of members of that OFFPOV defined family household?

Apologies if this has been raised before, I haven’t found an answer to my question elsewhere.

Your understanding is correct; Census determines poverty status independently for each family within a household. These families consist of an individual and anyone else related by birth, marriage, or adoption who resides with them. Importantly, unlike in the IPUMS created family variable FAMUNIT, unmarried partners are not considered members of their partner’s family. OFFPOV uses family size, total family income (FTOTVAL), and official poverty thresholds to determine poverty status. This page provides an instructive example of how poverty status is determined for a representative family.

Identifying each family presents a bit of a challenge. The best method might be using a combination of SERIAL and FTYPE (as well as YEAR and MONTH if you’re analyzing multiple samples), however you will need to pay attention to the type of family listed. Primary families and related subfamilies within the same household are treated as a single Census family. You can count the number of occurrences of FTYPE = 1 or 3 to arrive at the size of these types of families. Nonfamily householders and secondary individuals are simple to identify as they form their own one-person families. That is, family size = 1 if FTYPE = 2 or 5. Unrelated subfamilies are the most difficult as there may be multiple subfamilies within a household. In this case, you may also use OFFTOTVAL to identify members of the same subfamily as it is unlikely that the incomes of multiple unrelated subfamilies are equal. You can use CUTOFF to check your work as families of the same size should have the same cutoff values controlling for whether the householder is under 65.

Hi Ivan,

Thanks for your detailed response, I really appreciate your insight.

I was wondering whether using a combination of SERIAL and FAMID (using YEAR and MONTH for multiple samples) would help solve the problem of uniquely identifying families? My understanding of SERIAL is that it uniquely identifies each household, and that FAMID then uniquely identifies each family in a given household. Combining the two would allow me then to create an identifier that points to a specific family.

Is my understanding here correct or am I better off pursuing the method you outlined using FTYPE?


Hi John,

You are correct; FAMID is the family identifier that you’re looking for. It combines primary families and related subfamilies (FTYPE 1 and 3) and also identifies unrelated subfamilies, nonfamily householders, and secondary individuals as separate families per Census definitions. The method you outlined will allow you to uniquely identify families across households and samples.