HHTYPE variable, undetermined category

I’m having a hard time understanding the HHTYPE variable, particularly category 9 “HHTYPE could not be determined.” My understanding is that this variable is created by IPUMS to mirror a var that Census created, and that this undetermined category is largely made up of household heads with unmarried partners.

Are there some cases when an unmarried household head, living with an unmarried partner and a related child are considered family households, and classified either as single male, single female, or married couple households?

This part of the documentation is throwing me off: “ Family households with no spouse present include household heads of all marital statuses except married, spouse present (see MARST). Households where an unmarried partner is present are classified as family households only if there are other persons in the household who are related to the household head by birth, marriage, or adoption.” Then, there is a warning on another page of the codebook saying “With the inclusion of unmarried partners in the [SPLOC] pointer variable, HHTYPE ends up being set to 9 “HHTYPE could not be determined” in 1990-onward samples if a householder is linked to an unmarried partner."

Thank you in advance!

Your understanding of the purpose of HHTYPE and the 9 category are correct. The documentation on the description and comparability tabs for HHTYPE are at odds with one another; the user caution on the comparability tab (which notes that HHTYPE Is set to 9 if the household is linked to an unmarried partner) is generally accurate (with a few exceptions described below). The IPUMS USA team is planning to do a systematic review of variables related to family definitions in the next year or so; this will likely involve deciding how to handle HHTYPE for unmarried partners. In the short term, I will request that the IPUMS USA team update conflicting documentation.

Broadly, HHTYPE classifies households based on marital status (MARST) and relationship to householder (RELATE); previously, MARST and RELATE were lock-step with SPLOC (the probable spouse of a person as per the IPUMS family interrelationship variables). However, IPUMS recently updated our family interrelationship variables to include non-married partners when assigning SPLOC; this change had cascading effects for assigning HHTYPE. By including unmarried partners as part of SPLOC, we arguably indicate that these cases should be considered family households; however, that is at odds with the original description and intent of this variable. As an interim step, these households were all classified as “HHTYPE could not be determined” to allow the IPUMS USA team to decide if HHTYPE should reflect our updated family interrelationship definitions or prioritize the household type concept for unmarried partners as defined on page 316 of the 2000 PUMS codebook.

In 1990 only, households with unmarried partners (as per RELATE) are not necessarily classified as a 9 in HHTYPE. Details are below:

  • HHTYPE = 1
    • Households where the householder has an assigned spouse location (SPLOC>0) and a marital status of “Married, spouse present” regardless of the presence of an unmarried partner
    • These cases occur because HHTYPE prioritizes marital status over the RELATED value of the person’s spouse (specifically because the person reports being married with a spouse present)
  • HHTYPE = 9
    • Households where the householder has an assigned spouse location, the householder’s marital status value is something other than “Married, spouse present,” and the household contains an unmarried partner
    • This accounts for the bulk of unmarried partner’s HHTYPE values
    • In these cases, marital status does not override the information in RELATED (because they do not report being married with a spouse present)
  • HHTYPE not equal to 9 or 1
    • There a handful of unmarried partners who don’t fall into either of the above categories of HHTYPE
    • These are households that contain more than one unmarried partner

I suspect the difference between HHTYPE in 1990 and 2000-forward is driven by changes to Census Bureau editing procedures for relationship values that were not in place in 1990. I am sharing information on the ACS editing procedures and edits to the 2000 Census.

1 Like