Journal Issue: Marriage and Child Wellbeing Volume 15 Number 2 Fall 2005
Why Same-Sex Parenting Is Hard to Study
This body of research grew partly out of court cases in which lesbian and gay parents (or co-parents) sought to defend or obtain custody of children.10 Many researchers approached the subject with a sympathetic or protective attitude toward the children and families they studied. Critics have accused researchers of downplaying differences between children of gay and straight parents, especially if those differences could be interpreted unfavorably-—a charge that has been debated in the field.11 We will not enter that debate here, beyond noting that the best defense against bias is always to judge each study, whatever its author's motivation, critically and on its merits.
More significant, we believe, are the daunting methodological challenges that the researchers faced, especially at first.
Difficulty Finding Representative Samples
Perhaps the most important such challenge is that researchers have no complete listing of gay and lesbian parents from which to draw representative samples (probability samples, as researchers call them). To find study participants, they have often had to rely on word-of-mouth referrals, advertisements, and other recruiting tools that may produce samples not at all like the full population of gay and lesbian parents. All but one of the studies we examined employed samples composed of either totally or predominantly white participants. Almost all the participants were middle- to upper-middle-class, urban, well educated, and “out.” Most were lesbians, not gay men. Participants were often clustered in a single place. It may be that most same-sex parents are white, relatively affluent lesbians, or it may be merely that these parents are the easiest for researchers to find and recruit, or both may be partly true. No one knows. Absent probability samples, generalizing findings is impossible.
Small Sample Sizes
Gay- and lesbian-headed families can be difficult to locate, and funding for this research has been sparse.12 Those factors and others have forced researchers to deal with the challenge of small samples. Most studies describing the development of children raised in gay or lesbian homes report findings on fewer than twenty-five children, and most comparative studies compare fewer than thirty children in each of the groups studied. Other things being equal, the smaller the number of subjects in the groups studied, the harder it is to detect differences between those groups.13
The question is often not just how well same-sex parents and their children fare, but compared with whom? Should a single lesbian mother be compared with a single heterosexual mother? If so, divorced or never married? Should a two-mother family be compared with a two-biological-parent family, a mother-father family headed by one biological parent and one stepparent, or a single-parent family? It all depends on what the researcher wants to know. Identifying appropriate comparison groups has proved vexing, and no consistent or wholly convincing approach has emerged. Many studies mix family forms in both their homosexual and heterosexual groups, blurring the meaning of the comparison being made. Some studies do not use comparison groups at all and simply describe children or adults in same-sex households. Some, in fact, have argued that comparing gay and straight families, no matter how closely matched the groups, is inappropriate inasmuch as it assumes a “heterosexual norm” against which same-sex parents and their children should be judged.14
As we noted, families headed by same-sex parents are structurally very different from one another. That fact presents researchers with another challenge, because studies are most accurate when each of the groups being examined or compared is made up of similar individuals or families. When the pool of potential subjects is small, as it is for same-sex parents, assuring within-group homogeneity is often difficult. Thus some studies use “mixed” groups of lesbian-headed households, yielding results that are difficult to interpret. For example, partnered lesbians are often included with single lesbians, with all called “single” by the author; children who live both in and outside the home are discussed as a single group; children born into homes that originated both as heterosexual marriages and as lesbian households are included in the same sample; and separated and divorced women are mixed with nevermarried women and called “single.” In at least one of the studies reviewed, children of transsexuals and lesbians, children who are both biological and adopted, and parents who are both biological and adopters are treated as a single group.
Another challenge is to gauge how well children are faring. Few studies collect data from the children directly, and even fewer observe the children's behavior-—the gold standard for research of this kind, but more expensive and time-consuming than asking parents and children to evaluate themselves. Some studies use nonstandardized measures, while others use either measures with poor reliability and validity or measures whose reliability and validity were either not known or not reported.
Another measurement issue arises from the sometimes dated content of the measures used. In one 1986 study, for example, dressing in pants and wanting to be a doctor or lawyer were considered masculine for girls, and seeking leadership roles was considered a display of dominance.15 Those classifications look rather quaint today.
To some extent, researchers can compensate for heterogeneous samples and nonequivalent comparison groups by using statistical methods that control for differences, particularly in studies with larger samples. Not all studies have done so, especially in the era before today's advanced software made statistical work considerably easier. Some studies thus did not perform appropriate statistical analyses when that was possible. Others did not report the direction of the significant relationships that they found, leaving unclear which group of children fared better. Most failed to control for potentially confounding factors, such as divorce stress or the status of a current relationship with a former partner.
Putting the Research Challenges in Perspective
This is an imposing catalog of challenges and shortcomings, and it needs to be seen in context. The challenges we describe are by no means unique to the research on same-sex parenting, and neither are the flaws that result.16 Studying small, hard-to-locate populations is inherently difficult, especially if the subject pool is reticent. One of us, Meezan, has been conducting and reviewing field research on foster and adoptive families since the 1970s; he finds that the studies reviewed here are not under par by the standards of their discipline at the time they were conducted.