Thoughts on the Regnerus Study
Matthew Martin 3/26/2013 08:30:00 PM
Before I comment on the paper itself, I want to make a general point on principle. As a general ethical principle, no social science research should ever be used to justify the disparagement of anyone's basic human rights. To do so defies the Nuremburg code and all research ethics standards since. Moreover, to even conduct or publish such research with the foreknowledge that the results would be used disparage basic human rights is inherently unethical. There is literally no empirical result Regnerus could have found that would be in any way relevant to the question of whether LGBT couples should have equal rights. This is also the view that the Social Science Research editor James D. Wright, who published the Regnerus study, has expressed.
One thing that surprised me about the Regnerus study is that it was not as methodologically sloppy as I was led to believe by various media reports. There are, in fact, some respects in which the study actually does make valuable contributions to the research literature. For example, the study contributes the first nationally representative population-based survey of young adults (the NFSS refered to below). That does two things: first, it avoids a lot of the selection bias that pervades many of the existing datasets on LGBT households, and second, it provides us with reports from the child (after he grows up into a young adult), rather than the typical self-reports by the parents themselves.
That said, there are still methodological problems in how Regnerus conducts his analysis of the data. But, a lot of the hype around the paper has less to do with these flaws, per se, than it does the fact that anti-gay groups are just misrepresenting Regnerus's findings. In his own words, straight out of the paper
"There are several things the NFSS is not. The NFSS is not a longitudinal study, and therefore cannot attempt to broach questions of causation...It does not evaluate the offspring of gay marriages, since the vast majority of its respondents came of age prior to the legalization of gay marriage in several states. This study cannot answer political questions about same-sex relationships and their legal legitimacy."And from Regnerus' article in slate
"Let me be clear: I’m not claiming that sexual orientation is at fault here, or that I know about kids who are presently being raised by gay or lesbian parents. Their parents may be forging more stable relationships in an era that is more accepting and supportive of gay and lesbian couples."So, this paper is not about gay marriage, not about whether gay parents should be able to have kids, and not even about whether having gay parents affects child outcomes. It is just a descriptive paper looking at the demographics of children of parents who had a homosexual romantic relationship at some point, devoid of any policy implications or causal claims of any kind.
Now here are the facts. In the data that the paper uses, almost none of the respondents grew up in an LGBT household. In fact, of the 236 respondents whom the author defines as the children of gay parents, only 38 of them lived in an LGBT household for more than three years. That means that at most, only 16% of those actually grew up in an LGBT household. (Update: only two respondents grew up in a two-parent same sex household. So that's 0.8%, not 16% of those whom Regnerus defines as LGBT households). Almost all of the rest actually lived with their heterosexual father or mother after their biological parents separated. Because the incidence of homosexual preferences is exogenous (statistically speaking), this actually sets the paper up very well for doing a causal inference of the impact of divorce and separation on child outcomes (for serious wonks, think instrumental variables regression). Unfortunately, it tells us very little about the impact of having gay parents on child outcomes, because almost none of the respondents lived with their gay parent(s).
We should view the Regnerus study mostly as nothing more than an exploratory description of a potentially very useful dataset. It gives us a snap shot of the state of LGBT families, and what it shows ought to be cause for alarm--widespread and pervasive discrimination of LGBT people and disparagement of LGBT rights has left what would otherwise be very successful families in terrible disarray. Social norms have pushed LGBT individuals into heterosexual relationships, especially in the most conservative parts of the country, which only increases the likelihood of future divorce, which ultimately harms children. Compounding the injury, it appears that almost none of the LGBT parents are allowed to keep their children when they eventually settle down with the person (of the same sex) they love.
This is a bit of speculation, of course, because the study makes no attempt at causal inference. But what I see in the paper is powerfully suggestive evidence that what we need today is expanded marriage and parental rights for LGBT people everywhere, and that increasing acceptance of LGBT relationships would result in unambiguously better child outcomes because it would decrease the likelihood of children growing up in dysfunctional, divorced, broken households.
And as a final aside, I want to clear up some misconceptions. First, I've noticed a lot of anger towards Elsevier, the journal's publisher Elsevier--such anger is well placed but for the wrong reasons. We should be angry at Elsevier for profiting millions by restricting public access to research conducted by publicly-funded academics, while neither compensating the researchers nor providing any service of any measurable value. But, Elsevier is not involved at all int he decision of what to publish in its journals--that decision is up to the Journal editor, in this case James Wright, who in turn acts on the recommendation of the referees he selects to peer-review the article. In this case, it seems that the referees Wright chose may not have been very knowledgeable about the subject or research methods. Second, the journal in which this was published, Social Science Research, is honestly not very good. Not so terrible that anyone should be embarrassed by publishing there, but still not good enough to be highly selective in which articles it publishes. An expert should feel free to take results out of Social Science Research as they see fit, but the non-expert public should definitely give much better weight to the many studies in, say, the American Sociological Review, (which has an impact factor more than three times as high) which use actual causal inference models to show that having gay parents has no effect on child outcomes.