Friday, July 13, 2012

Witchhunt Under Way At University of Texas?

I don't know whether this study is actually defective or not.  The investigation under way by the University of Texas suggests to Instapundit that coming to the "wrong" conclusion could be the source of the investigation.
I don’t know if the study’s any good or not — but I’m positive that if it had found that children of gay couples do better, there would be no such inquiry, regardless of who had funded it. And the point here is mostly to warn other researchers that it’s just better not to look into certain subjects, or to come up with the wrong conclusions if you do.
It is worth reading the study.  The author points out that previous studies of children raised in homosexual homes have suffered from both sampling problems (non-random methods of acquiring test subjects) and comparisons to other less than optimal child-rearing situations:
Much early research on gay parents typically compared the child development outcomes of divorced lesbian mothers with those of divorced heterosexual mothers (Patterson, 1997). This was also the strategy employed by psychologist Fiona Tasker (2005), who compared lesbian mothers with single, divorced heterosexual mothers and found “no systematic differences between the quality of family relationships” therein. Wainright et al. (2004), using 44 cases in the nationally-representative Add Health data, reported that teenagers living with female same-sex parents displayed comparable self-esteem, psychological adjustment, academic achievement, delinquency, substance use, and family relationship quality to 44 demographically “matched” cases of adolescents with opposite-sex parents, suggesting that here too the comparisons were not likely made with respondents from stable, biologically-intact, married families. 
This is apparently the first really large, random study, involving 3000 adults, and concludes that children raised by both their biological parents are considerably better off than those raised in not only homosexual families, but also in heterosexual divorced families.
Besides being brand-new data, several other aspects about the NFSS are novel and noteworthy. First, it is a study of young adults rather than children or adolescents, with particular attention paid to reaching ample numbers of respondents who were raised by parents that had a same-sex relationship. Second, it is a much larger study than nearly all of its peers. The NFSS interviewed just under 3000 respondents, including 175 who reported their mother having had a same-sex romantic relationship and 73 who said the same about their father. Third, it is a weighted probability sample, from which meaningful statistical inferences and interpretations can be drawn. While the 2000 (and presumably, the 2010) US Census Integrated Public Use Microdata Series (IPUMS) offers the largest nationally-representative sample-based information about youth in same-sex households, the Census collects much less outcome information of interest. The NFSS, however, asked numerous questions about respondents’ social behaviors, health behaviors, and relationships. This manuscript provides the first glimpse into those outcomes by offering statistical comparisons of them among eight different family structures/experiences of origin. Accordingly, there is much that the NFSS offers, and not just about the particular research questions of this study.
The author points to the problem of self-selection of respondents, and how this may produce a very atypical sample.  The NFSS study is different in this respect:
The data collection was conducted by Knowledge Networks (or KN), a research firm with a very strong record of generating high-quality data for academic projects. Knowledge Networks recruited the first online research panel, dubbed the KnowledgePanel®, that is representative of the US population. Members of the KnowledgePanel® are randomly recruited by telephone and mail surveys, and households are provided with access to the Internet and computer hardware if needed. Unlike other Internet research panels sampling only individuals with Internet access who volunteer for research, the KnowledgePanel® is based on a sampling frame which includes both listed and unlisted numbers, those without a landline telephone and is not limited to current Internet users or computer owners, and does not accept self-selected volunteers. As a result, it is a random, nationally-representative sample of the American population. At last count, over 350 working papers, conference presentations, published articles, and books have used Knowledge Networks’ panels, including the 2009 National Survey of Sexual Health and Behavior, whose extensive results were featured in an entire volume of the Journal of Sexual Medicine—and prominently in the media—in 2010 (Herbenick et al., 2010). 
There are some great advantages to such a study, and one of them is that it presents a more typical sample of kids growing up in homosexual homes:
These distinctions between the NFSS—a population-based sample—and small studies of planned gay and lesbian families nevertheless raise again the question of just how unrepresentative convenience samples of gay and lesbian parents actually are. The use of a probability sample reveals that the young-adult children of parents who have had same-sex relationships (in the NFSS) look less like the children of today’s stereotypic gay and lesbian couples—white, upper–middle class, well-educated, employed, and prosperous—than many studies have tacitly or explicitly portrayed. Goldberg (2010, pp. 12–13) aptly notes that existing studies of lesbian and gay couples and their families have largely included “white, middle-class persons who are relatively ‘out’ in the gay community and who are living in urban areas,” while “working-class sexual minorities, racial or ethnic sexual minorities, sexual minorities who live in rural or isolated geographical areas” have been overlooked, understudied, and difficult to reach. Rosenfeld’s (2010) analysis of Census data suggests that 37% of children in lesbian cohabiting households are Black or Hispanic. Among respondents in the NFSS who said their mother had a same-sex relationship, 43% are Black or Hispanic. In the NLLFS, by contrast, only 6% are Black or Hispanic. 
Table 2 is what got the author in trouble.  It shows a variety of situations for these adult children ("Currently married", "Family received welfare growing up", "Currently on public assistance", "Thought recently about suicide").  Unsuprisingly, children of IBF (intact biological families) well outperformed most other situations, both heterosexual divorced and homosexual relationships.  (Not all differences are statistically significant; look carefully for the bolded values.)

One particular data point confirms what I have read elsewhere: children of IBF were much less likely to be in the "Ever touched sexually by parent/adult" category: 2%, vs. 23% from lesbian mother homes, 6% from gay father homes, 3% for adoptees, 10% for those whose parents divorced late, 12% for those raised in stepfamilies, 10% for single parent heterosexual upbringing.  (Bold indicates statistically significant difference from the IBF children.)

The study does not claim, nor do I, that gay and lesbian parents are going after their own kids.  The problem is that the instability of non-IBF homes creates opportunities for men who are not the biological father to get access to kids who otherwise might not be so available.  In addition, kids growing up in fatherless homes tend to look for approval from adult men, and this is easily manipulated by the Jerry Sanduskys of the world.

Those who want to see this study as some sort of conservative campaign to vilify homosexuality will need to explain why the study found that children growing up with lesbian mothers had generally much more obviously destructive outcomes than those growing up with gay fathers.  You would think if this was a "draw the curves, then find the needed data points" exercise, both would be similarly criticized.


Josh said...

The critiques are somewhat compelling: very few of the surveyed same-sex-children sets involve the child actually living in the same household as gay or lesbian parent and his or her lover for even three years (and almost none of the 'children of gay male couples' lived with a gay male pair for long at all), and then compared to only those children that lived with (married?) biological parents for their entire lifetime. Indeed, many of the children lumped into the "child of gay parents" group had spent very little time with his father after said father's sexuality came out, and the way of identifying which families had a gay, lesbian, or bisexual parent is not particularly good. Interesting information about the trauma of divorce, the terrible results of modern child custody law, or a number of other metrics, but without actually digging into them it's comparing apples to aluminum and claiming you're comparing apples to oranges. The datasets comparing like fields could be more useful and compelling (the random-level and are probably the best in the field today, and I'd love to see whether they show different results on some of the studies Regnerus criticizes at the start of his work), but the study seems to go out of its way to obfuscate those numbers.

((There a few other internal criticisms. A self-rating survey technique that neither queries nor proxies for alexithymia is terribly stupid, even if not an uncommon type of stupid.))

That doesn't really justify an investigation, though, at least not one into the study's author rather than the peer review process. I've seen much worse stretching of data before, and worse from UT-Austin in specific.

Clayton Cramer said...

That few of the children lived with gay fathers for any but the briefest of periods might explain why the various problems seemed more common among the children of lesbian mothers than children of gay fathers.

One of the points that the article makes well is that many of the existing studies purporting to show that gay parents are just as good or even better than IBF parents involve non-random samples and very small sample sizes (which makes subtle differences impossible to identify as statistically significant). This is a widespread problem with almost any form of sampling; people that volunteers for studies may be atypical for that reason.

For example, Kinsey's studies of human sexuality relied on very atypical populations: people willing to talk about their sexuality in the 1950s in the American Midwest were, to be blunt, going to be unusually open about sexuality. Some of his sample groups included mental hospital inmates, prisoners, and runaways. Some critics of Kinsey's numbers suggest that these alone explain the high percentage of homosexuals that he found, much less the rather primitive methods that he used to weight his raw data to conform to the U.S. population.

Josh said...

I'm not sure anything explains the child abuse, outside of possibly noise from the (very small) sample size of the "lesbian family" set. The children of gay fathers set not living with a pair of gay male adults means only that they're not good predictors for comparison to intact biological families, but a good portion lived with single gay fathers for sizeable periods of time, and a lot of the remainder went to single parent biomother households for some time. There's enough of a split that either explanation -- abuse by a gay parent or abuse because of a lack of a present male parent -- should at least show something similar to the rate among stepfamilies. And either explanation comes across as weirdly lackluster in contrast to parents divorcing late (after the child reached 18) correlates with a fivefold increase in sexual abuse by parents/adults -- ruling out time travel, you're still stuck with weird causative tracks.

The base information is very interesting, and I'd love to crunch the numbers myself. There's no evidence I've seen that the raw data has been manipulated, and that's why it does give some interesting results. At the same time, the presentation of the data leaves a lot of places to mislead.

I'll note that Kinsey didn't find terribly high amounts of homosexuality; the concept as we know it today was somewhat controversial, and Kinsey himself didn't particularly like the distinction. He found high amounts of same-sex physical interaction or arousal, which is a rather separate matter, and one that remains both when you clean the weird sample selection out of his data and from other researchers. Remains fairly constant despite increasing willingness to talk about sex in mainstream society. It doesn't terribly mean much -- Kinsey's methodology is about the only one that would call someone gay for one drunken same-sex kiss followed by three years of watching het-only stuff -- but that's a separate critique, and one that more rightfully needs to be focused on the folk misrepresenting Kinsey's research.

Clayton Cramer said...

I rather doubt that gay fathers are going to be molesting their children, both because of a fathers genuine concern for his children, and because especially in the period which this covers, even the suspicion of molestation by an openly gay father would in most states have led to custody changes so fast that your head would spin.

Roger said...

This study is good evidence that IBF families work best. That should not be any surprise, as that has been conventional wisdom for millennia.