July 18, 2001
Recently, the Human Rights Campaign, an alliance working for lesbian, gay, bisexual and transgender equal rights, released reports claiming that Census Bureau data from the 2000 Census shows a dramatic increase in the number of households consisting of homosexual couples. Some reporters and editorialists have had a field day proclaiming hundreds of percentage points of increase in the number of homosexual couples in some states. Homosexual advocates have declared that the increase in homosexual households reflected more tolerant attitudes toward homosexuality. The Human Rights Campaign cited these alleged increases as evidence of “incredible progress.” Those jubilant claims subsequently were cut short when the Census Bureau published a notice: “As a result of [technical] changes in the processing routines, estimates of same-sex unmarried partners are not comparable between the 1990 and 2000 Census.”
As it turns out, there are major problems with the data on homosexual households:
First, no meaningful comparisons can be made between the 1990 and 2000 Census as it relates to the number of same-sex partners because the data was tabulated using different procedures. The differences in numbers of same-sex “married-couple” households in the 1990 and 2000 Census was due in large part to the way the Census Bureau interpreted the data.
In the 1990 Census, when same-sex “marriage” was not a legally permissible designation anywhere in the nation, couples of the same sex who identified themselves as a “spouse” were counted inconsistently by the Census Bureau-sometimes as a “married couple;” other times as siblings or roommates. Also, if one of the partners was more than fifteen years younger than the other, they might have been reported as child and parent.
In the 2000 Census, the Census Bureau used the 1996 Federal Defense of Marriage Act (DOMA) to invalidate same-sex spouse responses. The Census Bureau turned DOMA on its head by explaining that the invalidations as well as “social and legal aspects” of the problem necessitated putting the “spouse” designations of same-sex couples into a new category, “unmarried partners.” For the 2000 report, if respondents of the same sex designated themselves as a “spouse,” the Census Bureau counted them in the “unmarried partner” tally.
In short, there is no way that the Census Bureau data can provide an objective measure of an increase in the number of gay couple households in the U.S. since 1990-a fact bluntly stated twice in the Bureau’s unusual notice.
Second, the Census Bureau erred in eliminating the marital status question from the 2000 questionnaire-the so-called “short form” questionnaire for the 2000 Census did not even include a question about marital status. This bureaucratic maneuvering precludes any meaningful interpretation of the census data regarding marriage and eliminates the possibility of comparisons with previous or future years and makes trends difficult to determine and predict when relying solely on Census data. Without uniform definitions of categories and consistent allocation of responses, the quality of the data is seriously degraded. We should note that the Census Bureau specifically stated: “estimates of opposite-sex unmarried partners [co-habitation data], however, were not affected by these editing procedures and changes, and are comparable between the two censuses.”
Third, percentage increases can be misleading. Percentage changes are especially dramatic when dealing with small numbers. With the most accurate estimates of the total homosexual population at 2% to 3%, real number changes are more meaningful than percentage changes. For instance, adding two pennies to one results in a 200% increase that distorts the relatively insignificant real number change.
Perhaps the most important lesson of this story is the illustration it provides about the use and misuse of statistical data in general, and about the use of Census Bureau data in particular. It is always a good idea to inquire about the wording of the questions that were asked as well as how the information was obtained from respondents and how the surveys were processed. The Census Bureau utilizes volunteers who simply ask people questions about the characteristics of their households as they fill out official questionnaires. Then, they turn in the questionnaires to be tabulated by Bureau officials. The accuracy of the data the volunteers collect depends to a great extent upon the honesty of the people questioned and those who record their answers. The actual meaning that should be attached to the responses depends on how the responses are interpreted (coded) and tabulated by the Census Bureau.
Census data are still considered generally reliable and are indeed the most comprehensive source of information we have about the composition of American households. Nevertheless, it is important to keep in mind that any data source has its limitations as a source of truth. Social science statistics are certainly useful, and they can help shed light on a great many questions about human behavior. But they are always subject to interpretation, and they must always be considered in the context of how the data were collected, categorized and tabulated.