Census1990: Married HH Head + Unmarried Partner, Mistake?

#1

Howdy,

In the 1990 Census, there appear to be cases of households with a married HH head, the HHhead’s spouse, and then an unmarried partner. Is this correct / faithful to the raw data? I count about 3,600 HH’s that have this feature. Below are three examples (1 with a spouse, 2 without).

Thank you!
–Luke

Example #1: Serial==2876; numprec==4
pernum==1 : age(50), marst(1), sex(1), related(Head)
pernum==2 : age(47), marst(1), sex(2), related(Spouse)
pernum==3 : age(23), marst(6), sex(1), related(Child)
pernum==4 : age(20), marst(6), sex(2), related(Unmarried Partner)

–> To me, I would expect pernum(4) to be the partner of pernum(2).

Example #2: Serial==4278; numprec==4
pernum==1 : age(30), marst(1), sex(1), related(Head)
pernum==2 : age(4), marst(6), sex(2), related(Child)
pernum==3 : age(3), marst(6), sex(1), related(Child)
pernum==4 : age(31), marst(1), sex(2), related(Unmarried Partner)

–> To me, I would expect that pernum(4) should be the spouse of pernum(1).

Example #3: Serial==7840; numprec==2
pernum==1 : age(45), marst(1), sex(1), related(Head)
pernum==2 : age(37), marst(1), sex(2), related(Unmarried Partner)

–> To me, I would expect that pernum(2) should be the spouse of pernum(1).

— EDITS:
I put “relate” but really the variable was “related”

#2

Additional Issue: this is related, but different.

Background: I am forming family units so am ultimately interested in matching children to parents (and I typically agree more with the Census than IPUMS, so do not use IPUMS nchild).

IPUMS is creating different subfamilies based on related==1114 “unmarried partner”, so that related==301 “child” is being assigned by default to a different subfamily than related==101 “Head”. It seems odd that children are assigned to the subfamily of the “unmarried partner”, since the “unmarried partner” might not be biologically related to the children.

EDIT: I was wrong about “default” – see example 3. But now it seems somewhat inconsistent which get assigned.

The variable nchild does not seem to be affected, and, as nchild is “social children”, it seems consistent with the spirit of the variable. When there are children of the “unmarried partner” but likely not the HH head, the children have relate==1260 “non-relative” and are in the same subfamily as the “unmarried partner”.

For example 1: serial==246; numprec==4
pernum==1 : related(101), subfam(0), nchild(2), age(32), sex(2)
pernum==2 : related(301), subfam(1), nchild(0), age(13), sex(2)
pernum==3 : related(301), subfam(1), nchild(0), age(07), sex(1)
pernum==4 : related(1114), subfam(1), nchild(2), age(37), sex(1)

–> It seems odd to put pernum(2,3) in a different subfamily than pernum(1), and it seems particularly odd to default the HH head’s children to the unmarried partner’s subfamily rather than the HH head.

For example 2: serial==3549; numprec==3
pernum==1 : related(101), subfam(0), nchild(1), age(49), sex(2)
pernum==2 : related(1114), subfam(1), nchild(0), age(38), sex(1)
pernum==3 : related(1260), subfam(1), nchild(1), age(9), sex(2)

–> It seems appropriate to put pernum(2,3) in a different subfamily than pernum(1) because pernum(3) is likely not a biological child of pernum(1).

For example 3: serial==617905; numprec==4
pernum==1 : related(101), subfam(0), nchild(2), age(47), sex(2)
pernum==2 : related(301), subfam(0), nchild(0), age(18), sex(2)
pernum==3 : related(301), subfam(0), nchild(0), age(19), sex(1)
pernum==4 : related(1114), subfam(1), nchild(2), age(53), sex(1)

–> It seems appropriate to put pernum(2,3) in the same subfamily than pernum(1) because pernum(2,3) are likely biological children of pernum(1).

#3

On the first question: Yes, this is what the raw data say. The RELATED values are provided directly by respondents and therefore characterize their relationship. The meaning of the “unmarried partner” category, however, significantly changes over time. As is stated on the comparability tab: “This category changes considerably in meaning across census years. […] In 1990, 2000, and the ACS, this category is split into ‘Housemate or roommate’ and ‘Unmarried partner’–the latter of which is supposed to distinguish those with a “close personal relationship” with the householder.”

On the second question: This is a known error. As you suggest the observations in SERIAL==246 should not be considered subfamilies, since they are either assigned to be the spouse to the household head or children of the household head. The IPUMS USA Team is currently working on fixing this issue.

#4

Thanks for the quick response and thorough answer!