Some caution about Census data on same-sex couples:

Yesterday I posted some results from a new Census study comparing married same-sex couples and married opposite-sex couples, and comparing both to unmarried same-sex couples. The study concluded that in many significant ways — including likelihood to be raising children, income, home ownership, education, and race — gay and straight married couples were very similar, and unlike unmarried gay couples. The study also furnished evidence for some surprises, like the possibility that lesbian couples might be less likely to get married (or consider themselves married) than gay male couples.

But there's a potential problem in at least some of the Census numbers, which are inconsistent with some of the work done by demographers who have studied same-sex couples. Gary Gates, a noted demographer and researcher at UCLA (and a gay-marriage supporter), writes in an email to me:

One of the take-aways from the [Census] report is that married same-sex couples look quite a bit like their different-sex counterparts. That may very well be true, but one reason for the similarity is that it's quite possible that a very large portion of the married same-sex couples are, in fact, different-sex married couples who miscoded their sex. I've attached a paper (presented at the same session of the Population Association of America conference as the Census paper) that describes the difficulties in interpreting the married same-sex couple data.

From other work I've done, we know that married same-sex couples are 2-to-1 female and that women are more likely than men to be in a partnership. This isn't very consistent with the Census findings. Our analyses suggest that the sex miscoding problem among married different-sex couples creates more male same-sex couple miscodes than female. That could explain the Census findings.

There are several other findings that are not consistent with information we have about differences between cohabiting same-sex couples who are or are not in legally recognized relationships. For example, in a paper I published recently in Demography (with Christopher Carpenter), we show that those in registered domestic partnerships (in CA) ["RDPs"] have higher income and education levels and are more likely to be white than those who are not registered. These are the same patterns we see among heterosexual couples (comparing married v. unmarried) and contradict the Census findings. We also find no evidence of higher rates of child-rearing for those in RDPs in men and modest evidence of differences among women. Granted, RDP and marriage are not the same, but folks should be very cautious in interpreting the Census findings.

I think it's a very positive step that the Census released an analysis of the same-sex spouses. But it's just a first step. Much more work is needed to better understand who the married same-sex couples are and how many are miscodes.

If Gates is right about the coding problems, the miscoding would have skewed the results in favor of similarities since opposite-sex couples would have been included in the "same-sex" data. So gay and straight couples may be alike in many of the ways the Census Bureau suggested, but the new Census data do not necessarily support that hypothesis. A lot more work is needed, including more work based on the 2010 Census itself. In the meantime, modesty and caution about this new Census data are in order — more modesty and caution than I used yesterday.

UPDATE: For some interesting historical background on an especially noxious Census error, see here.

Related Posts (on one page):

  1. Some caution about Census data on same-sex couples:
  2. New Census study comparing gay and straight married couples:
rosetta's stones:
Focus on the real data concerning gays in the military. That's an avenue that's not been fully explored, I'd think. I gotta believe there's enough there to make a difference, here and from other countries.
6.20.2009 7:10pm
Randy R. (mail):
The Census isn't like a poll. It strives for an accurate counting, rather than a representative counting. If it were representative, it would likely be closer to Gary Gates analysis. Because it strives for accuracy, but can't achieve total accuracy (it's difficult to say you've accounted for 100% of the people), some will inevitably be left out. Those left out may skew things differently than if only a representative sample were taken.

So yes, it doesn't mean the Census Bureau is wrong, but it doesn't mean it explains everything. It's just one for bit of information to be added to the pile.
6.20.2009 11:20pm
James Gibson (mail):
I did say your use of the data seemed forced. Instead, it was even worse: premature. Errors in Census data, or just plain government document, is an old problem going back centuries. My grandfather in 1930 was identified as Elenore instead of Elmore. If it wasn't for the mark as Male it really would have been stupid. In the same regard, it was recently reported that a same sex couple got married in New York even though the state presently doesn't recognize it. It turns out one of the partners had a document that miss-identified him as a her.

Census takers tend to rush: I should know I was one in the 1990 census. They also tend to make spelling errors and have trouble printing long names in the space provided on the forms. People then make things even worse by the name they give for themselves or their children. Sometimes they give their middle name (like my dad), or their nickname, because people know them better by it then their given first name. I have two males ancestors whose full name was so long that the census takers truncated the first two (or three) names to only initials leaving only the last name fully represented. It made finding them in the census a test of my searching skills.
6.21.2009 2:59am
Putting Two and Two...:
miss-identified him as a her

As opposed to mister-identifying her as a him


Sorry, couldn't resist.

But, I wonder, does this mean we can get our equal rights if we just prove we're smarter than those hetero couples who can't even fill out simple paperwork?
6.21.2009 3:16am
Lymis (mail):
Analysis of this sort of data, especially in states where there is not legal protection from things like getting fired for being gay, needs to include at least the awareness that there are solid reasons for some people not to choose to self-report.

Married people with kids are already out, both legally and socially. The registered with some government entity somewhere, and are likely to be out to their children's schools and so on.

People with more education are likely to be working in jobs that have more real or perceived protections.

Two men with kids are going to be more socially obvious to passersby than two women, so even those who would prefer to be closeted are less likely to.

Responsible analysis wouldn't conclude things like "gay people are more educated" but rather "educated gay people are more likely to self-report." Using census data to conclude, for example, that there are more white gay Ivy League graduates than black gay rural Southerners would be problematic.

This will be iterative. Each census that asks these questions will get more accurate, and people will be more willing to self-report as conditions in their areas and in their lives improve.
6.21.2009 10:06am
Pro Natura (mail):
6.21.2009 11:09am