When the U.S. Census Bureau released 2010 census numbers earlier this year, the percentage of same-sex couples was reported generally higher in Central Texas and throughout the country. We wrote about the change in the Statesman, including maps showing the change from 2000-2010 in each Texas county and within places in Central Texas. New estimates released this week put some of those details we reported under a cloud, but the general premise that Travis County ranks highest in the state for same-sex couples still holds true.
This change was a significant increase compared to recent American Community Survey estimates, so the Census Bureau took a closer look at the results. What they found were errors recorded by census takers going door to door while coding “relationship to householder” and the gender of the person were enough to overinflate same-sex couples. Since the same-sex couples segment is so much smaller than the opposite-sex couples segment, even a small percentage of error exaggerated the same-sex couples segment, according to the Census Bureau.
Because of this discrepancy, the bureau released new estimates that show fewer same-sex couple households in Texas than the 2010 census data first suggested. The percentage of same-sex couple households in Texas went from .756 percent in the 2010 Census to .493 percent in the revised numbers, a quarter of a percent drop, or down from 67,413 households in the 2010 census to 43,045 in the new estimate. Comparisons of percentage of same-sex couples across counties is still reasonably accurate since the chance of error was constant, writes Martin O’Connell and Sarah Feliz in their paper Same-sex Couple Household Statistics from the 2010 Census , so our overall premise that Travis County leads the state in same-sex couples should not change. But specific numbers within a geographic area, like in our map published online with the story, are subject to the reporting errors. In most cases, the number of same-sex couples are overinflated.
The Census Bureau suggests we can use the 2010 ACS 1-year estimates to compare areas of at least 65,000 people, which would include Hays, Travis and Williamson counties. However, since the ACS is not a full count, but a survey of only small number of people, the margin of error gets higher as you get to smaller geographies. (Example: In Hays County, the survey counted zero “male householder and male partner” households, but the margin of error was +/- 294 households. Other recent 3-year and 5-year estimates, which use larger sample sizes, have had between 40 and 105 couples for that data point.) We’ll be able to get a better comparison when we have the 2010 ACS 3-year estimates due in October. Those estimates should provide a big enough sample to compare, but not include the bad reporting mechanism present prior to 2008.
The form problem found in the 2010 door-to-door counts was actually discovered during the 2007 ACS and was corrected for the 2008 ACS and beyond. The Census Bureau was able to update the 2010 Census form sent by mail to American households, but was not able to correct in time the form used by enumerators going door to door, according to the bureau.
The actual form problem is interesting to look at, too. What the bureau found was when the boxes for male and female were stacked on top of each other, people were much more likely to mark it incorrectly, drawing threw one box into another, compared to when the gender boxes were side by side.
So how has the Census Bureau figured out their revised estimates? They went back to the door-to-door form results and compared the first name of the person to the gender marked, and then ran the comparison against the likelihood that the name was for a male or female. When they made changes based on the first name and reran the estimates for same-sex couples, the results were more like ACS results from the previous years. They provided these estimates at the state level earlier this week, but don’t plan to go any deeper geographically, said Robert Bernstein, public affairs specialist at the Washington D.C. office.