"‘Explosion' in the Gay and Lesbian Population?"

Was there an "explosion" in the number of gay and lesbian Americans during the last decade of the twentieth century? Anyone reading the newspapers in July 2001 would have answered yes. Headlines roared that the number of gay and lesbian households had "skyrocketed" (Gram 2001) and that the nuclear family was "fading" (The Gazette 2001) and "in meltdown" (Feder 2001). The evidence? These news reports focused on data from the 1990 and 2000 decennial censuses of the United States, which reported that the number of gay or lesbian couples living together increased drastically between 1990 and 2000.

For example, in the article "Jump in Reporting of Same-Sex Households in New York," writer Michael Hill reported that the census counted 13,748 same-sex partner households in New York State in 1990 and 46,490 such households in 2000—more than a threefold increase. Similarly, the number of gay households in the United States also jumped more than threefold, from approximately 145,000 to 600,000. In the article, Hill described a lesbian couple, Diana and Laurie Spencer, who felt comfortable living in a small town in upstate New York because of the feeling that as the number of gay households increased acceptance of gays would increase as well.

While gay activists found these data to be affirming, conservative critics cited the census data as "evidence" that the traditional family, and morality more generally, were "unraveling before our eyes" (Dobson 2001). Demographers and survey researchers, however, were less concerned with the political implications and more concerned with figuring out whether these statistics were correct. Most concluded that the number of gay households probably had not increased substantially. Rather, the purported increase reflected changes in how the U.S. Bureau of the Census counts and classifies American households and their residents.

Every ten years, the Census Bureau obtains a full roster of information on all persons who live in a given household. For each household, a "householder" or "head of household" is identified, and then all persons in the household are categorized in terms of their relationship to the householder. The Census Bureau also obtains information on the age and gender of all persons living in a household. Taken together, these data can be used to document patterns such as the number of grandchildren under age eighteen who are living with their grandparents or the number of adult siblings who are living together. However, the specific ways that the members of a household are described and classified have changed. In the 1990 census, if a person residing in a household reported being the "spouse" of the householder and being of the same sex as the householder, then census data analysts reclassified this person as a "boarder," or "roommate." In other cases, the Census Bureau recoded the person's gender, perhaps thinking that the gender was reported in error and that the couple was in fact an opposite-sex married couple (Beveridge 2007).

By the year 2000, however, that very same person—a person who indicated being of the same sex as the householder and being the "spouse" of the householder—would be classified as an "unmarried partner." Consequently, is it possible to definitively conclude that the number of gay households increased between 1990 and 2000, or would it be more accurate to conclude that the number of gay or lesbian households was undercounted in the 1990 census?

Experts believe that the latter is true. Angry gay and lesbian advocacy groups shared their concern with the Census Bureau, spurring the Census Bureau to publicly acknowledge its questionable analytic strategies on its Web site. The Census Bureau also urged data users to recognize that direct comparisons between the 1990 and 2000 censuses would be impossible. The Web site said, "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....We believe 2000 estimates of this category are better estimates than those produced in 1990" (U.S. Bureau of the Census 2001). This incident illustrates an important issue in sociological research methods, particularly survey research methods. If the precise question or method used to measure a particular concept changes over time, then it is very difficult to determine whether social change is really happening or whether a social pattern has been stable but is simply being counted or described in a new way.