Inductive reasoning and empirical evidence regarding the fortunes of children from same-sex unions

8 Oct

When it comes to family structures, Americans (and American public policy) can take (and has taken) one of three positions: it can legally recognize but not prefer/encourage a particular family structure; it can legally recognize but remain indifferent or non-preferential towards them all; or it can not legally recognize a particular family structure at all.  Historically, traditional or natural marriage/family structures have been legally recognized and encouraged in the law or policy-programs of civil government.  Other family structures have been legally recognized but not necessarily legally preferred/privileged/encouraged (e.g., single-parenthood; “open” marriages; cohabitation).  As family structural arrangements involving divorced parents has become normalized, in many ways it has become legally recognized and treated in public policy with non-preferential indifference.  Still others, like polygamy and in most states same-sex marriage, have not been legally recognized at all.

Child-custody law and policy is a case in point.  Courts, social workers, and other public officials are presented an array of family structures and, at least for now, they have not remained indifferent or non-preferential between them.  All things being equal, parents in traditional family structures are not exclusively recognized but they are generally preferred to other family structures in determining custody.  In the Welfare Reform Act, signed by Bill Clinton, both traditional and single-parent homes were legally recognized but only the former was preferred or encouraged explicitly in the reform plan (each state had to take measures to  reduce out-of-wedlock births).

Which approach will be used to address same-sex marriage remains to be seen, though legal recognition appears to be winning the day.  It is interesting, however, given the clear precedent and practice, that few in the debate consider the middle-option these days (legal recognition of same-sex unions without undermining legal preference for traditional marriage generally as a matter of public policy).  It seems that no one is satisfied with the civil-union position, even though it could be thought of as a kind of compromise.  Such a position would essentially preserve traditional marriage as the ideal, preferred, privileged position for the purposes of procreation and child welfare, while extending marital benefits to same-sex couples.  Much as we do with regard to divorced or single parents seeking adoption, it would be treated as less than ideal while traditional homes are treated as ideal in law and practice.  But both sides seem more interested in having same-sex marriage go the way of polygamy (legally condemned) or legal equivalence with traditional marriage (legal indifference between traditional marriage and SSM).

And yet, the evidence that family structures are pivotal, even foundational, in child-welfare has mounted and continues to mount.  Children simply fare better on virtually every measure of child-welfare, in traditional homes (raised by married biological parents), ceterus paribus.  This has been true regardless of which alternative family structure we are talking about (divorced, open marriage, cohabitation, single-parent, etc.).  Given this evidence, it is rather odd that it has in just a few years time become those who promote traditional marriage that are seen as irrational dinosaurs when they doubt that children raised in same-sex households will fare any better than all of the other alternative candidates to traditional homes.  Maybe they are just using inductive reasoning, which is the traditional basis of scientific inquiry (i.e., they reason: all the other alternative forms consistently fare poorly for children; why should the next one, whatever it happens to look like, be any different?).  Indeed, while it’s perhaps a bit too early to draw hard empirical conclusions (certainly in America where SSM is so new), if we look to the Canadian experience, the skeptics about the inconsequential outcomes of SSM appear to be justified in a recent study (excerpt):

There is a new and significant piece of evidence in the social science debate about gay parenting and the unique contributions that mothers and fathers make to their children’s flourishing. A study published last week in the journal Review of the Economics of the Household—analyzing data from a very large, population-based sample—reveals that the children of gay and lesbian couples are only about 65 percent as likely to have graduated from high school as the children of married, opposite-sex couples. And gender matters, too: girls are more apt to struggle than boys, with daughters of gay parents displaying dramatically low graduation rates.

Unlike US-based studies, this one evaluates a 20 percent sample of the Canadian census, where same-sex couples have had access to all taxation and government benefits since 1997 and to marriage since 2005.

While in the US Census same-sex households have to be guessed at based on the gender and number of self-reported heads-of-household, young adults in the Canadian census were asked, “Are you the child of a male or female same-sex married or common law couple?” While study author and economist Douglas Allen noted that very many children in Canada who live with a gay or lesbian parent are actually living with a single mother—a finding consonant with that detected in the 2012 New Family Structures Study—he was able to isolate and analyze hundreds of children living with a gay or lesbian couple (either married or in a “common law” relationship akin to cohabitation).

So the study is able to compare—side by side—the young-adult children of same-sex couples and opposite-sex couples, as well as children growing up in single-parent homes and other types of households. Three key findings stood out to Allen:

children of married opposite-sex families have a high graduation rate compared to the others; children of lesbian families have a very low graduation rate compared to the others; and the other four types [common law, gay, single mother, single father] are similar to each other and lie in between the married/lesbian extremes.

Employing regression models and series of control variables, Allen concludes that the substandard performance cannot be attributed to lower school attendance or the more modest education of gay or lesbian parents. Indeed, same-sex parents were characterized by higher levels of education, and their children were more likely to be enrolled in school than even those of married, opposite-sex couples. And yet their children are notably more likely to lag in finishing their own schooling.

The same is true of the young-adult children of common law parents, as well as single mothers and single fathers, highlighting how little—when you lean on large, high-quality samples—the data have actually changed over the past few decades. The intact, married mother-and-father household remains the gold standard for children’s progress through school. What is surprising in the Canadian data is the revelation that lesbian couples’ children fared worse, on average, than even those of single parents.

The truly unique aspect of Allen’s study, however, may be its ability to distinguish gender-specific effects of same-sex households on children. He writes:

the particular gender mix of a same-sex household has a dramatic difference in the association with child graduation. Consider the case of girls. . . . Regardless of the controls and whether or not girls are currently living in a gay or lesbian household, the odds of graduating from high school are considerably lower than any other household type. Indeed, girls living in gay households are only 15 percent as likely to graduate compared to girls from opposite sex married homes.

Thus although the children of same-sex couples fare worse overall, the disparity is unequally shared, but is instead based on the combination of the gender of child and gender of parents. Boys fare better—that is, they’re more likely to have finished high school—in gay households than in lesbian households. For girls, the opposite is true. Thus the study undermines not only claims about “no differences” but also assertions that moms and dads are interchangeable. They’re not.

Read the whole summary of the study here

5 Responses to “Inductive reasoning and empirical evidence regarding the fortunes of children from same-sex unions”

  1. J. Palmer October 8, 2013 at 2:37 pm #

    “…children of gay and lesbian couples are only about 65 percent as likely to have graduated from high school as the children of married, opposite-sex couples.”

    Considering the children of same-sex couples are adopted, wouldn’t a more valid comparison be with the adopted children of opposite-sex couples?

    Since “Adopted children are more likely than biological children to have special health care needs, current moderate or severe health problems, learning disability, developmental delay or physical impairment, and other mental health difficulties,” a comparison to non-adopted children seems like stacking the deck.


    • thereformedmind October 8, 2013 at 10:00 pm #

      Your observation is addressed in the article.


  2. thereformedmind October 10, 2013 at 5:13 pm #

    I took down a comment on this blog post from a Mr. Scott Rose. Unfortunately, Mr. Rose littered his commentary with personal attacks and name calling. I will not allow this blog to become a platform for hate speech and religious bigotry (ironically coming from one preaching tolerance and acceptance).

    However, when he was not attacking and accusing and demeaning the characters and speculating about the motives of professional researchers whose research reaches conclusions incommensurate with Rose’s worldview or ideology, he manages to make a worthy point. It is true that the bulk of studies on this subject (same-sex parenting outcomes for children) do not reach the conclusions that Allen or Regnerus have reached (SSM, like divorce, produces outcomes unfavorable to child welfare, all things else being equal). However, the vast majority of those studies (I’ve looked at this myself) are small-n experimental designs. What they have in terms of internal validity (low contamination of alternate factors) they greatly lack in external validity (ability to generalize to larger samples of populations). This is just the nature of small-n experimental designs (typically employed in psychology studies). Non-experimental designs, featuring large-n samples, are the best places to look for this issue since we want to know what the general pattern is across large populations. I’m not the only one to point this out. Consider the signed statement by this group of scholars:

    In fact, as others have pointed out (, of the prior research, it appears that only the Rosenfield study (2010) featured a large-n sample. Allen’s study is also a large-n sample and in the article and elsewhere, both scholars dispute the methodology of the other (without resorting to the ad hominen attacks Mr. Rose seems to prefer. It is possible to have this debate both at the popular and scholarly level without such behavior, but to do so, we’ll have to tune out the likes of Mr. Rose.

    I plan on reading both the Allen and Rosenfield studies soon for comment. In the meantime (Mr. Rose), consider something Allen wrote (in a far different tone) about this whole debate some time ago, and then ask yourself who is really being blinded by bias and hate?


  3. thereformedmind October 10, 2013 at 5:15 pm #

    I failed to include this breakdown of the extant literature review as well:


  4. thereformedmind October 10, 2013 at 5:56 pm #

    I’ve just read the Allen, Pakulak, and Price (2013) critique of Rosenfield’s study, in which they use his data, replicate his findings and reevaluate his results. Here’s the link (but it’s behind a pay wall):

    The study is very good methodologically, and open and honest about each step taken (including a note regarding any appropriate concerns). Yes, the biggest concern is that they included all children of SSM couples regardless of how long they lived in the home and whether the child was adopted or their own. BUT THAT IS NOT ALL THEY DID. They had multiple models, varying in degree of sample restrictiveness (some included both restrictions; some only one or the other). In only one scenario (the most restricted, using Rosenfield’s 5 year living in household and “own child” rule) did the comparison between SSM couples fall out of statistical significance. In every other scenario, the children performed as they do in every other alternative family structure on most outcomes (this one, BTW, is limited to normal educational progress; we will have to wait for researchers to see if differences exist on other outcomes). And that is less well relatively speaking.

    Even more, Rosenfield failed to included odds ratios and standard errors beside his regression coefficients so a true sense of the magnitude of the confidence intervals could be seen. Yes, the highly restricted model (which reduces the sample of children by 50%) does not pass the standard 95% confidence interval. But it does using a slightly more forgiving one. (I’ll have to check to see if either researchers are using a one or two-tailed test; a good case can be made for one, I would think, giving extant literature on other family structures; this may indeed make statistical significance even at the .05 level easier to obtain).

    Perhaps one of the most interesting things they did was compare SSM children with every other alternative family structure (something Rosenfield didn’t do). Here they find that SSM structures can’t be distinguished in terms of outcomes from single-parenthood, which we simply know in the literature is associated with worse outcomes on nearly everything.


Comments are closed.

%d bloggers like this: