Most gay-parenting studies are long on bias and short on hard data

National Review 14 June 2012
Douglas W. Allen is the Burnaby Mountain Professor of Economics at Simon Fraser University in British Columbia
I am a Canadian economist who has worked on family issues in Canada and the U.S. for the past 26 years. Although I’ve mostly studied matters of divorce, custody, child support, and the general institution of marriage, for the past few years I’ve been working on series of empirical projects related to same-sex marriage. I’ve been using a special data set in Canada that is large (over 300,000 individuals) and random (with weights), that directly identifies sexual orientation, and that was designed by Statistics Canada. In the process of working on same-sex marriage I have read almost every study conducted on same-sex parenting. I say all of this because, unlike most people who have commented on the recent Regnerus study, I’m a qualified outsider to the U.S. debate and perhaps can provide some (relatively) neutral assessment.

The study published by Professor Mark Regnerus this week certainly has some flaws, and many of the comments made about it have some merit. However, as a matter of intellectual honesty, it needs to be recognized that virtually all the studies of same-sex parenting that have been conducted thus far fall far short of any standard of scientific testing.

Of the 50-plus such studies done in the past 15 years, the vast majority come to the same conclusion: Children of gay parents perform at least as well as children from heterosexual families; there is no difference in child outcomes based on family structure.

For several reasons, this literature is unlike anything else within social science. First, it partly arose from, and was strongly influenced by, legal cases in which lesbian mothers were denied custody of their children on the basis of their sexual orientation. Second, for the most part it has been written by individuals with strong personal worldviews who sympathize with those studied. Third, the focus of the literature is often on “soft” measures of child and family performance that are not easily verifiable by third-party replication, and that differ substantially from measures used in other family studies. One of the odd characteristics of this literature is the lack of consistency of measures across time. Subsequent studies seldom test for measures that were used in previous studies. Fourth, the data and procedures used in the studies are never made available online in order for other scholars to replicate findings. And finally, almost all the literature on gay parenting is based on weak designs, biased samples, and low-powered tests.

The result is a nascent literature that falls far short of standard social-science research. At its best, the literature contains interesting exploratory studies that raise provocative questions and make interesting observations. At its worst, it is advocacy aimed at legislators and judges — which may explain why, despite its weak scientific nature, the literature is characterized by strong recommendations for policy and legal changes to family regulations.

The bias of the same-sex-parenting literature has been recognized by individuals within and outside this literature (indeed, in the same issue of Social Science Research as the Regnerus study, Loren Marks has provided another critique of this literature). Ironically, the common complaint about Regnerus — that he compares apples to oranges — is valid about practically every study that finds no difference between homosexual and heterosexual families. In the latter, biased samples of high-income, highly educated, self-selected lesbian parents are compared to random samples of opposite-sexed parents.

If the Regnerus study is to be thrown out, then practically everything else in the field has to go with it.

I think Regnerus needs to be applauded for what he did and didn’t do. He tried to use a random sample; he developed many hard measures of outcomes; and he is making all the data and procedures available for others to sift through. Inadvertently, he is going to draw attention to the failures of other studies in terms of their design and methodology, and he has demonstrated how difficult it is to find a large sample of this elusive population. He also didn’t make a lot of unjustified claims in his study. He was careful to note that he made no case for causality, and that his findings may or may not be related to the same-sex aspect of the adult relationship. He didn’t take his results and announce a series of policy recommendations. He has simply called into question the claim that there is no difference.

Others are working on this very issue, and soon better studies will be published. In my own work, I’m also finding differences in behavior and in child outcomes. Given how small the population of same-sex parents is, given how many different channels children might take to find themselves in a family with two parents of the same sex, and given how much data it takes to sort through all of these issues, the bottom line is this: We’ve got a long way to go before we can answer the question: Are children better off, the same, or worse off in same-sex families compared to intact biological families?

The political contest that is going on in the U.S. over same-sex marriage is not helping the social science. It took almost 40 years for academics to figure out the effect of no-fault divorce on divorce rates (not to mention all the other areas of life no-fault divorce influenced). With same-sex marriage and parenting, the issues are much more profound and more difficult to measure. Rushing the work or, worse, pushing research claims beyond what the studies justify, is bad social policy. This goes for both sides of the debate.

Written by