Census drops couple estimate by 40%

The Census Bureau on Tuesday released a new estimate for the number of same-sex couple households in the United States in 2010, and it is almost 40 percent fewer than it actually counted. But it’s 80 percent higher than the number counted in the 2000 Census. And the number of same-sex couple households represented less than one percent of all households in the United States.

These were some of the conclusions reported by Census Bureau officials in a telephone press conference with reporters September 27. The data is based on the Bureau’s own analysis of responses to household types on the 2010 Census from 50 states, the District of Columbia, but an analysis deemed credible by the Williams Institute.

According to the data, the Census Bureau counted 901,997 same-sex couple households in the 2010 survey, but it believes only 646,464 of those are really same-sex couples. The other 255,533, said Census Bureau official Martin O’Connell, are heterosexual couples whose Census forms inadvertently misidentified the gender of one of the partners.

O’Connell said the Census Bureau based its “preferred” estimate on a complicated analysis of the first names of the partners in same-sex couples. If both partners in a couple had very clearly male or very clearly female names—like Thomas and John or Deborah and Elizabeth—the same-sex couple designation was considered very reliable. But if one partner or spouse in a same-sex couple had a name like Leslie or Jessie, the same-sex designation was considered unreliable and the couple was re-classified as a heterosexual couple.

That’s because the Census Bureau discovered there were a great deal more unreliable gender-name designations on forms that were filled out by in-person “enumerators”—Census representatives who visited households that failed to mail back their standard Census form. The form used by these in-person enumerators, said O’Connell, were problematic and contributed to the likelihood that the enumerator might check off the wrong gender box while filling out the form.

“We only used names where 95 percent or more of the time that name was associated with being either a male or female,” said O’Connell, “so we tried to limit those [uncertain] areas by basically taking a very conservative aspect by only removing from the same-sex couple universe those names which had an extremely high probability of being of a sex that was opposite of what they reported.”

Responses reporting gender were considered most accurate in 12 states and the District of Columbia. Those 12 states included Massachusetts, Vermont, Connecticut, California, Oregon, Washington, Nevada, Arizona, New Mexico, Colorado, Minnesota, and New York. They had “inconsistent name-sex reporting” in 6 to 26 percent of same-sex couples.

Eight states had the worst accuracy (between 38 and 50 percent inconsistencies), including Arkansas, Alabama, Mississippi, West Virginia, Montana, Wyoming, and South and North Dakotas.

The correction process had a dramatic effect on the number of same-sex couple households counted in many individual states. For instance, California’s original data count was 125,516 same-sex couple households, but its corrected, or “preferred,” estimate totaled only 98,153 – a loss of 27,363, or almost 22 percent. Massachusetts’ number dropped from 26,049 in raw numbers to 20,256 in corrected numbers – a loss of 5,793 couples, or 22 percent. Florida dropped from 65,601 to 48,496 –a loss of 17,105 couples, or 26 percent. New York dropped from 65,303 to 48932 – a loss of 16,371 couples, or 25 percent. Georgia dropped from 29,844 to 21,318 –a loss of 8,526 couples or 28.5 percent. Michigan dropped from 21,782 to 14,598 –a loss of 7,184, or 33 percent. Texas dropped from 67,413 to 46,401 –a loss of 21,012 couples, or 31 percent.

The Census Bureau used data from Texas to illustrate how well first names corresponded with the gender identification and, therefore, how reliable the designation of same-sex couples held up. Of 31,763 households designated originally as a male-male couple, only 14,439 (45 percent) qualified as including “highly likely male” —that is, both partners or spouses had first names that are almost certainly male names, such as Thomas or John.

O’Connell said the problem with the in-person form wasn’t discovered until late in 2009— too late for the Census Bureau to correct and print out new forms in time for use in 2010.

Gary Gates, a respected expert in LGBT demographics at the Williams Institute, said the Census Bureau’s procedures are both “credible” and “substantially more accurate” than the numbers released on a state-by-state basis over the summer.

“I was part of a team that identified this problem with Census Bureau data collection procedures more than seven years ago,” said Gates, “and it is a positive step to see the Bureau release these new estimates.”

Of the 646,464 that the Census Bureau counts as same-sex couples, 514,735 (80 percent) were unmarried partners and 131,729 (20 percent) were married spouses.

In 2000, the Census Bureau’s “preferred estimate” of same-sex couple households was 358,390, meaning the number of same-sex couple households increased by 80 percent in the past decade. Of the 358,390 same-sex couple households identified in 2000, 88 percent identified as unmarried partners while 12 percent identified as married. Same-sex marriage did not become available anywhere in the world until 2001 (in the Netherlands). It became available in the U.S. (in Massachusetts) in 2004.

Using the uncorrected counts, the number of same-sex couple households increased by almost 52 percent between 2000 and 2010. There were 594,391 same-sex couple households counted in 2000, compared to 901,997 counted in 2010.

In the 2010 preferred data, the state with the highest percentage of same-sex couples in relation to the total number of households in the state was Vermont, with 1.091 percent. Massachusetts was second, with 1.023 percent, and California was third with 0.998 percent.

Gates said the 2010 effort, while credible, is only a “temporary solution.”

“Ultimately, they have to change the way the questions are asked in order to minimize these types of errors in the data,” said Gates. “Canada and the UK use procedures known to substantially minimize the problems observed in U.S. data. The Census Bureau has tested these possible revisions and should move quickly to change current surveys and collect more accurate data.”

Top Ten Same-sex Couple States
The following states have the highest percentage of same-sex couple households compared to the total number of households in the state. The number represents the percentage of same-sex couple households in the state based on a corrected, or “preferred,” estimate calculated by the Census Bureau from Census 2010 data. The District of Columbia, a city that has a population that exceeds that of several states and which is often regarded as a state-like territory, had a percentage of 1.929. New York came in 12th with 0.892 percent. Florida ranked 14th with 0.884 percent. Georgia ranked 16th with 0.832 percent. Illinois ranked 28th with 0.671 percent. North Dakota ranked 50th with 0.396 percent. The U.S. average was 0.773.

1. Vermont 1.091
2. Massachusetts 1.023
3. California 0.998
4. Oregon 0.986
5. New Mexico 0.984
6. Delaware 0.979
7. Maine 0.970
8. Hawaii 0.933
9. Washington State 0.927
10 Nevada 0.926

One Response to Census drops couple estimate by 40%

  1. S says:

    So if the in-person enumerators were this inaccurate at checking off which sex people were, aren’t there also likely to be a lot of couples incorrectly identified as opposite-sex when they’re really same-sex? Did the Census also go through all the opposite-sex couples and perform the same name/gender/95% test to convert any of those from opposite-sex to same-sex?

    Or was the sole result of this exercise an elimination of same-sex couples, and meanwhile many of us are still not being counted due to being incorrectly labeled us as opposite-sex couples? Seems like a 40% error rate could go both ways.

Leave a Reply