Assessing the Australian Study
National Review Online 6 June 2013
“Children of same-sex parents are happier and have healthier familial relationships than their peers with parents in straight relationships,” or so says what is purported to be the world’s largest study on the children of same-sex parents. As always in this domain, I read the early media input about this with immediate interest — and a bit of skepticism, too, given the glowing, confident enthusiasm displayed online. So I went fishing for more information about the interim report — not readily locatable yet — and about the Australian Study of Child Health in Same-Sex Families (ACHESS) in general, and found information about its methodology here. To summarize (and quote from) it:
Initial recruitment will involve convenience sampling and snowball recruitment techniques. . . . This will include advertisements and media releases in gay and lesbian press, flyers at gay and lesbian social and support groups, and investigator attendance at gay and lesbian community events. . . . Primarily recruitment will be through emails posted on gay and lesbian community email lists aimed at same-sex parenting. This will include, but not be limited to, Gay Dads Australia and the Rainbow Families Council of Victoria.
The ACHESS, about which I’m sure we’ll hear a great deal over the next few days and weeks, is thus a lot like the National Longitudinal Lesbian Family Study (NLLFS), except that it’s larger and more recent in its generation. I realize that 500 cases is not a number to scoff at, and that such populations are a small minority to begin with. But until social scientists decide to do the difficult, expensive work of locating gay parents through random, population-based sampling strategies — and ones that do not “give away” the primary research question(s) up front — we simply cannot know whether claims like “no differences” or “healthier and happier than” this or that group are true, valid, and on target. Why? Because nonrandom samples are not a representative reflection of the population as a whole, but rather an image of those who actively pursue participating in the study (for whatever reason, which may matter). Who knows — the ACHESS sample of parents and children could be just like the average gay or lesbian household in Australia. I have my doubts, but it’s an unanswerable question.
While noting the possibility of bias in such a sampling approach, the ACHESS architects assert that this is “not possible to overcome due to the hidden nature of the same-sex attracted population in Australia.” How hidden are they if the sampling design simply floated emails and flyers and attracted over 500 participants? “Hidden” here is shorthand for difficult to locate randomly. Difficult, but not impossible. Difficult equals expensive. But a random sample design is the gold standard in large-scale social science. And in a politically charged environment such as gay parenting, the public would do well to demand nothing less than the best-quality research designs. Snowball sampling — where motivated friends ask their own friends to participate — doesn’t cut it.
Another cause for some healthy skepticism is the same reason I suggested that it was time to retire the NLLFS. Simply put, its participants are likely very aware of the political import of the study topic, and an unknown number of them probably signed up for that very reason. As a result, I’m just not sure I trust their self-reports, which may be subject to considerable “social desirability bias,” or the tendency to portray oneself on surveys as better than one actually is. Again, it’s unknowable here. But I think the temptation to do so in this sample, and on this topic, could be elevated. All the more reason to do a random study that doesn’t advertise its intentions beforehand.
All these concerns are why the survey I oversaw, the New Family Structures Study (NFSS), elected to talk to the children after they had grown up, to skip the parents entirely to ensure a more independent assessment, to not broadcast our key research questions in the title or initial screener questionnaire, and to locate them randomly in a large population-based sample. “Junk science” it is not. But its initial results, as well as follow-up analyses in response to criticism, certainly differ from those in Australia making the headlines today.
Finally, a note about children born via assisted reproductive technology (ART) or adoption, which no doubt make up a significant share of the ACHESS. A central reason that could set this non-random group apart from “average” others — even if data were collected randomly and independently — is the expense involved in acquiring them. If you’re talking about ART or adoption, you’re talking about parents with notable means — money and commitment. That is, when children are not born in the usual way — via the products of vaginal sexual intercourse, half of which are unplanned (but not necessarily unwanted) — it implies an initial outlay of wealth to adopt or donor conceive that is not consonant with the average heterosexual couple’s life or experience. (A random sample of average Australian parents appears to have been the comparison group in this “interim” report.) In other words, there are no unplanned or unwanted pregnancies among monogamous gay and lesbian couples, and that matters. But “diminished kinship” remains — meaning at least one parent lacks a biological tie to the child — and that matters for child development, too, at least on average.
In the era the NFSS investigated, ART births appear to have been much less common. Nor did its nationally representative, population-based sample reveal very many stably coupled same-sex households with children. Things could be different today. But I stand by the NFSS — and always will — because it captured the social reality of the era it studied: the growing-up years of people ages 18 to 39 today. The ACHESS and the NLLFS capture some sort of reality — perhaps subject to significant social-desirability bias — among a self-selected elite of participants who were actively recruited to take part in a study loaded with possible political import. But given the limitations outlined here, what we can learn from it about the overall population of same-sex households with children — which is what the media headlines convey — is pretty modest. We learn what’s possible, not necessarily what’s probable.
– Mark Regnerus is an associate professor of sociology at the University of Texas at Austin and a research associate with the university’s Population Research Center. He is also the principal investigator of the New Family Structures Study.