METHODOLOGI CAL DETAI LS AUGUST 2007 Danielle MacCartney, Public Policy Fellow M.V. Lee Badgett, Research Director Gary J. Gates, Senior Policy Fellow United States Decennial Census, 2000 The primary data source for estimates about gay and lesbian households in the United States for the Census Snapshots published by the Williams Institute is the United States 2000 Decennial Census. Unless otherwise noted, we use the 5 percent Public Use Microdata Sample (PUMS) to determine the characteristics of same-sex unmarried partner households. The PUMS is a random sample of responses to the census long-form, which contains detailed information about all members of the household. I dentifying same-sex unmarried partners The census household roster includes a number of relationship categories to define how individuals in a household are related to the householder (the person filling out the form). These fall into two broad categories: related persons (e.g., husband/wife, son/daughter, brother/sister), and unrelated persons (e.g., unmarried partner, housemate/roommate, roomer/border, and other nonrelative). Since 1990, the Census Bureau has included an “unmarried partner” category to describe an unrelated household member’s relationship to the householder. If the householder designates another adult of the same sex as his or her “unmarried partner” or “husband/wife”, the Census counts the household as a same-sex unmarried partner household. These same-sex couples are commonly understood to be primarily gay and lesbian couples (Black et al. 2000), even though the census does not ask any questions about sexual orientation, sexual behavior, or sexual attraction—three common ways used to identify gay men and lesbians in surveys. Measurement issues Several measurement issues associated with the same-sex unmarried partner data could affect estimates describing the gay and lesbian population of the United States. First, to the extent that the census sample can be used to derive characteristics of gay and lesbian people, it is important to note that the sample is only representative of same-sex couples. Their characteristics may differ from those of single gay men and lesbians. For example, Carpenter and Gates (2007) find that white and highly educated gay men and lesbians are more likely to be partnered. Second, concerns about confidentiality may lead some same-sex couples to indicate a status that would not provide evidence of the true nature of their relationship. Other couples may believe that “unmarried partner” or “husband/wife” does not accurately describe their relationship or household structure. A study of undercount issues relating to same-sex unmarried partners in Census 2000 indicates that these were the two most common reasons that gay and lesbian couples chose not to designate themselves as unmarried partners (Badgett and Rogers 2003). It seems reasonable to believe that the census tends to capture same-sex couples who are more willing to THE WI LLI AMS I NSTI TUTE | CENSUS SNAPSHOT | METHODOLOGI CAL I SSUES | AUGUST 2007 1