Unexpected Same Sex Marriages in 1990 Census


I am forming Tax Units to run NBER TaxSim on, so I am grouping together married persons. I am using the 1990 Census 5%, so in this time period a married unit can only be a Husband + Wife (Man + Woman).

However, I can find instances where the SPLOC variable links two person with the same sex. Is this a coding mistake or an interpretation of the data by IPUMS? Or am I incorrect about the possibility of same sex marriage in 1990?

Pernum==3: RELATED==1115; AGE==63; MARST==1, Woman==0, SPLOC==4
Pernum==4: RELATED==1115; AGE==63; MARST==1, Woman==0, SPLOC==3

Pernum==2: RELATED==1115; AGE==27; MARST==1, Woman==0, SPLOC==3
Pernum==3: RELATED==1115; AGE==58; MARST==1, Woman==0, SPLOC==2

Pernum==2: RELATED==301; AGE==17; MARST==1, Woman==0, SPLOC==3
Pernum==3: RELATED==301; AGE==16; MARST==1, Woman==0, SPLOC==2

Pernum==3: RELATED==301; AGE==19; MARST==1, Woman==1, SPLOC==5
Pernum==5: RELATED==401; AGE==19; MARST==1, Woman==1, SPLOC==3

Any thoughts?

You are correct that same-sex couples cannot be consistently identified before the 2000 Census, however, same-sex couples can be probabilisticly identified in the 1990 Census. Although this category was refined by the time of the 2000 Census, the 1990 Census is the first year that the “unmarried partner” category exists in the detailed codes of the RELATE variable. This allows for the probabilistic identification of same-sex couples in the 1990 5% sample. More information about this detail is available on this page.