Same-sex parenting studies: the re-analysis scoreboard

by Carolyn Moynihan. May 22, 2015.

A study going the rounds of the newsrooms this week is said to debunk, discredit and otherwise disprove the findings of the New Family Structures Study carried out by University of Texas sociologist Mark Regnerus on the effects of same-sex parenting.

Authors Brian Powell and Andrew Cheng, sociology professors from Indiana University and the University of Connecticut, have reanalysed the NFSS data and reclassified the respondents who said they had lived with a lesbian mother or gay father for less than a year, as not “raised” by a same-sex parent (Regnerus seems to have used the term “raised” to summarise a range of exposure to same-sex parents) and even questioned the inclusion of those who spent less than four years in such a household. They then sorted the 236 subjects by the stability of their parents’ relationships and added a control for poverty in addition to Regnerus’ income control.

In other words, after eliminating a handful of genuine oddballs cases, they went on to do a different study, one that explained away almost all the 40 significant differences Regnerus found in favour of intact biological families.

Regnerus, who made his data available for researchers, has responded to this latest effort to stamp out dissent in the profession from the “consensus” that there is “no difference” between children raised by same-sex couples and those raised by married moms and dads.

In “Making differences disappear: the evolution of science on same-sex households” he compares what Powell and Cheng have done with his study to erasing the association between exposure to Agent Orange and cancer in Vietnam veterans by controlling for sustained exposure to heavy forestation in combat zones. In other words, it was being near forests sprayed by the chemical that gave them cancer, not the chemical itself. It was the mom’s divorce, not the fact that she had a lesbian partner for a while, that we should suspect for the young adult’s troubles later on.

Regnerus’ problem is not with reanalysis, which he says is a valid method in the long-term process of cleaning and clarification in any dataset of substantial size. But it’s not the same thing as setting out to obliterate significant associations actually in that data.

The lesbian studies with very few lesbians

Nor are studies finding problems with same-sex parenting the only ones with data that needs cleaning up. Four studies regarded as the strongest of those showing “no differences” for children from same-sex households have all used corrupted data, leading to invalid findings.

Between 2004 and 2008 Dr Jennifer Wainright and Dr Charlotte Patterson authored or led three studies based on US data from the National Longitudinal Survey of Adolescent Health (“Add Health”). These studies were the first to go beyond convenience samples of same-sex parents and use a representative population sample with sufficient statistical power to discern differences, if they existed.

Out of 20,000 adolescents in the 1994 Add Health data the researchers found 44 with lesbian mothers. They compared them with a matched group of young people with heterosexual parents on a number of measures: psychosocial well-being, school performance, and romantic relationships, delinquency, victimization and substance abuse. They concluded that “adolescents living with same-sex parents did not differ from that of adolescents living with opposite-sex parents” in any way that would disadvantage the former.

But these studies suffer from a fatal flaw. Owing to wrong coding in the original Add Health survey, 27 of the 44 children Wainright and Patterson counted as having lesbian parents were actually children with both a mother and a father in the home.

This is what Professor Paul Sullins of the Catholic University of America found when he studied the original data, as reported in his unpublished paper, “The Unexpected Harm of Same-Sex Marriage: A Critical Appraisal, Replication and Re-Analysis of Wainright and Patterson’s Studies of Adolescents with Same-Sex Parents.” It is not surprising, he says in a summary of his paper within a brief submitted recently to the Supreme Court, that there were no differences between the two groups of adolescents, given that they were largely from the same type of household. Obviously these studies cannot support the “no difference consensus”.

And the other gold standard study using corrupt data

The same applies to a fourth random sample study based on data from the US Census 2000, where many heterosexual partners were classified as same-sex couples, as pointed out by Dan Black and colleagues in a 2007 study published by the California Centre for Population Research. They estimated that at least 40 percent of the cases in the same-sex couples sample of the census “are actually different-sex married couples,” and concluded by warning researchers that “many of the inferences drawn from these data are incorrect.”

Still, in 2010 Stanford sociologist Michael J. Rosenfeld published a study of children’s progress in school based on the corrupt data, apparently unaware it was flawed. Sorting the data correctly invalidated his “no difference” findings, says Sullins. On the other hand it supported a 2013 Canadian study by economics professor Douglas W. Allen, who found that children living with gay and lesbian families in 2006 were about 65 percent as likely to graduate from high school compared to children living in opposite sex marriage families. Daughters of same-sex parents did considerably worse than sons.

Sullins’ own research showing emotional problems among children of same-sex parents found support from the Wainright and Patterson data once only clear cases of same-sex couples were identified. Reanalysis confirmed that anxiety was significantly higher for children with same-sex parents.

More importantly, it also indicated that marriage would not improve the experience of children in same-sex households. On the contrary:

Comparing the