The Census Bureau reduced the number of same-sex couples it counted in the 2010 decennial census. Originally, when numbers were released in August of last year, more than 900,000 couples were counted, but now there are only 646,000, which is a pretty big drop.
The Census Bureau claims that the distribution of same-sex couples has remained unchanged. It seems rather unlikely that the distribution remained unchanged. For this to be true, I think a few things would have had to happen. Since the problems in identifying couples had to do with identifying the sex of respondents,a uniform process of identifying sex would have to have been done over all recorded Census questionnaires. This part, particularly if computerized, is not that hard (in STATA-speak–replace gender=1 if firstname==”Patricia”, replace gender==0 if firstname==”Michael”, etc). But it also is necessary to have taken proportional numbers from each state or region, which seems like it would be difficult to maintain over such a large area as the US. For instance, if 5% of the same-sex couple households reported in August were in Colorado, then exactly 5% of that extra 30% would have to be misclassified Coloradans. And the same for every state. It just seems like regional differences in naming and willingness to self-report same-sex household status would be different enough over states and urban/rural to skew it somehow.
I know that the Census Bureau imputes a lot of values and also works very hard to be accurate in the face of a lot of problematic data, but for almost 30% of the sample to disappear and still have the same geographic distribution seems unlikely.
But maybe I’m not giving us or the Census Bureau enough credit.