Abstract: Despite claims that “no differences” exist between children whose parents had a same-sex relationship and children who were raised by their married biological parents, previous research cannot support such an assertion. Using a large, nationally representative dataset, a new study by sociologist Mark Regnerus finds that children whose parents had a same-sex relationship experienced more negative adult outcomes compared with children from intact biological families. The study has sparked a remarkably hostile and unscientific backlash—a backlash presumably motivated by the paper’s implications for the same-sex marriage debate. This backlash is regrettable because it undermines the health of public discourse on a subject of enormous significance—the institution of marriage—and challenges the integrity of social science inquiry in general.
In 2005, the American Psychological Association (APA) declared: “Not a single study has found children of lesbian or homosexual parents to be disadvantaged in any significant respect relative to children of heterosexual parents.” This sweeping “no difference” claim has been cited by proponents of same-sex marriage to support redefining marriage.
However, the implications of the APA’s conclusion are actually much more limited, due to the kind of research upon which it is based. Often forced to work with small and unrepresentative samples, social scientists have been unable to determine how average children with same-sex parents compare with average children raised by their married biological parents. Such generalizable conclusions based on samples representing the whole population are the kind of information that citizens should expect from social science that is used to support a significant policy proposal, but the previous literature cannot provide this information.
A June 2012 study in Social Science Research by University of Texas sociologist Mark Regnerus helps to shed some light on the issue. Using a nationally representative random sample called the New Family Structures Study (NFSS), it finds that a number of negative outcomes are associated with having a parent who was in a same-sex relationship compared with having two married biological parents.
The Regnerus study improves on prior methods and represents an important contribution to research on family structures. Three critical reviews of the study published in the same edition of Social Science Research hail it as such.
Elsewhere, however, the study has sparked a remarkably hostile and unscientific backlash—a backlash presumably motivated by the paper’s findings that undercut the “no differences” claim. The backlash is regrettable because it both undermines the health of public discourse related to a subject of enormous significance—the institution of marriage—and challenges the integrity of social science inquiry in general.
The New Family Structures Study shows that prior research is inadequate to support any assertion that it makes no difference whether a child was raised by same-sex parents. The results suggest both that there is a lot more to learn about how changing family forms can affect children and that social science evidence offers an insufficient basis for redefining marriage.
The Existing Literature
In social science research, failure to find evidence of a hypothesized effect does not automatically mean that the effect does not exist. The quality of the research involved, especially regarding the size and representativeness of datasets, helps social scientists to determine whether the hypothesized effects are truly nonexistent or merely undetectable with the statistical tools at their disposal.
Much of the past research on outcomes for children in same-sex versus heterosexual households lacks the ability to confidently rule out the possibility of differences when comparing averages among the broader population. In particular, the main challenge to research on the children of parents in same-sex relationships has been simply finding enough of them to analyze in the first place. Most existing datasets with detailed demographic data do not contain enough children of parents in same-sex relationships to conduct an informative analysis. For example, the widely used Add Health dataset contains only about 50 such children, despite a core sample of 12,105 adolescents.
Researchers have generally compensated by creating “convenience” samples—sets of respondents that are easily obtainable by the researcher but do not necessarily reflect the average characteristics of the population in question. For example, one technique for creating a convenience sample of parents in same-sex relationships is to advertise in homosexual-themed newspapers and magazines. Researchers then ask the people who answer the advertisement to recommend others who might be willing to participate. The next set of respondents is asked for more leads, and so on, creating a “snowball” sample that the researcher can then use.
It is not difficult to see how convenience samples of this kind can be unrepresentative of same-sex parents in general. People who are having poor experiences as parents may be less likely than contented parents to volunteer for a survey. Selection through snowball techniques also tends to produce samples that are relatively homogeneous, with prior same-sex parent studies often dominated by upper-class urban whites.
Obtaining an unbiased sample is a crucial aspect of social science research in general. Regardless of the subject matter at hand or the population being examined, large and representative samples are essential for drawing strong conclusions about a particular group.
In a literature review published in the same issue of Social Science Research as the Regnerus study, family studies professor Loren Marks detailed several other problems in the dataset construction in studies of children whose parents had same-sex relationships. These problems include small sample sizes, lack of comparison groups, and narrow sets of outcome measures. All of these objections are applications of standard research principles—not arcane technical points.
It is important to note, just as Regnerus does, that researchers producing these past studies were usually open about the limitations of their methodologies, and their work is still interesting and informative in certain ways. But it cannot tell us how the average children of same-sex parents compare in terms of stability and outcomes with the average children of married biological parents.
The Regnerus Study
To improve on prior methodologies seeking to compare children’s outcomes across household types, Mark Regnerus led an ideologically diverse team of researchers from multiple universities who advised on the design of the New Family Structures Study (NFSS). The NFSS features a sample of 2,998 adults between the ages of 18 and 39, with information from respondents about both their childhood experiences and their current circumstances as adults. The sample contains 175 respondents who reported that their mothers were in a same-sex relationship at some point during their childhood and 73 whose fathers were in such a relationship during their childhood.
Unlike much of the past research on the topic, these respondents are derived from a random population-level sample that is much more likely to reflect the average experience of children with a parent who had a same-sex relationship. Both Regnerus and his critics would like to have a larger number of such children to study, but the NFSS sample size does provide considerably more statistical power in detecting differences compared with most of the past research. Regnerus examined 40 different outcomes—many more than any previous study—and controlled for a variety of family circumstances.
Results of the Regnerus study reveal that having a parent who is or has been in a same-sex relationship is generally associated with more negative adult outcomes, especially when compared with adult children from intact biological families. For example, adults whose mother or father had a same-sex relationship have lower educational attainment than adults who grew up with their two married biological parents. They are also more likely to receive welfare, experience depression, smoke, and be arrested. These differences remain after controlling for a variety of other childhood circumstances, such as race, family income, and state of residence.
On 24 of the 40 outcomes after controls, Regnerus found statistically significant differences (meaning highly unlikely to have been due to random chance) between children whose mothers had same-sex relationships and children who grew up in intact biological families. Children whose fathers had a same-sex relationship were significantly different from children in intact families on 19 measured outcomes after controls.
What the Study Does and Does Not Say
As Regnerus makes clear, these results establish an association among family structure, parental relationships, and adult outcomes—not causation. The study does not by itself establish that having a parent in a same-sex relationship is a root cause of the differences in outcomes that Regnerus observed. However, it does suggest that such a causal mechanism is plausible and cannot be ruled out. The claim of no measured disadvantage for children with parents who have same-sex relationships cannot be justified by the existing research.
Despite what some media reports might suggest, Regnerus’s study draws no conclusions about how marriage should be defined. The report focuses on the data, not their implications for the political and legal debate. However, his study is significant because its findings discredit a popular argument for same-sex marriage: that it makes no difference whether children are raised by parents who had a same-sex relationship or by a married mother and father, an argument that the existing data cannot support.
Hopefully, future research that builds on the Regnerus study will use even larger samples with more control variables. Longitudinal study designs, meaning those that follow the same children over time, could be especially illuminating. Also interesting would be an exploration of whether outcomes for children with parents in same-sex relationships vary based on the birth cohort of the subjects (1972–1993), given that the adults profiled in the NFSS grew up at a time when such relationships were less publicly visible.
Engaging the Substance of the Study: Are Negative Outcomes Due Simply to Unstable Families?
One criticism leveled at the Regnerus study is that it does not limit its comparison to stable families headed by committed same-sex couples. Instead, Regnerus categorizes respondents based on their reports of having a parent who had a same-sex relationship—a much broader category that includes some parents who at one time were in heterosexual relationships. This has led some observers to argue that the Regnerus study is really capturing the impact of unstable family structures rather than the impact of having a parent in a same-sex relationship per se.
The importance of this objection has been overstated for several reasons. First, in a follow-up study, Regnerus separated respondents who lived with their mother and her same-sex partner from respondents who never lived with their mother’s same-sex partner. Compared with respondents from intact biological families, respondents who lived with their mother and her same-sex partner reported significantly different outcomes on 19 of the 40 measures after controls.
Second, in addition to the primary comparison group of respondents who were raised in intact biological families, the original study examined several other family forms. Although the differences were not nearly as great as compared with intact biological families, respondents whose parent had a same-sex relationship also generally fared worse than respondents with divorced or single parents.
For example, compared with children whose mothers had a same-sex relationship, both children with single parents and children with stepparents were less likely to receive welfare when growing up, more likely to be employed as adults, less likely to be depressed, and less likely to be arrested.
Importantly, the original study reports only the raw comparisons—not controlled for other family circumstances—when the control group is not the intact biological family. Investigating how children with parents in same-sex relationships compare with other family forms is an important avenue for further research, and the initial data suggest that differences may exist.
A third reason the Regnerus study is not merely capturing the effects of family structure is that same-sex parenting involves greater risk of instability relative to biological families insofar as one parent is biologically related to the child and one parent is not. Even if nothing about sexual orientation or same-sex gender dynamics affects the quality of parenting, outcomes associated with such same-sex step-families are likely to entail the same challenges and instability associated with heterosexual step-families.
In addition, emerging data indicate high divorce rates among same-sex couples. In Scandinavia, same-sex civil unions—essentially marriages in everything but name—have been legal for about two decades. After controlling for age, region, country of birth, education, and duration of the partnership, male couples in Sweden were 35 percent more likely to divorce than heterosexual couples, and lesbian partners were over 200 percent more likely to divorce. Whether the couples have children makes little difference in the relative rates.
The reasons for these higher rates of divorce are unclear. They could be due to same-sex gender dynamics, different social expectations for same-sex versus opposite-sex unions, or perhaps omitted control variables. In any case, researchers should continue to study the higher divorce rates closely to evaluate the effects of same-sex parenting.
Interestingly, Regnerus has said that he would gladly have included an analysis of children raised in stable two-parent same-sex homes, but there were not enough in his data. In fact, he found only two such households after screening over 15,000 participants. This fact alone suggests that stability in same-sex parenting is a legitimate concern, although it should be emphasized—as Regnerus emphasizes—that same-sex parenting is probably more common now than it was during the childhoods of those studied in the NFSS (born between 1972 and 1993).
Critics Avoiding the Substance of the Study
Scientific research should be evaluated strictly on its methodological merits, not on the political implications of the results. Regrettably, much commentary on the Regnerus study has failed to meet this basic standard. Rather than treat the substance of Regnerus’s study, many opponents have used exaggerated denunciations (“junk science”) and made baseless accusations of political bias and scholarly impropriety.
Beyond labeling the study “dangerous propaganda” and “appalling and irresponsible,” opponents have sought to discredit the author himself. An assistant editor at The New Republic called Regnerus a “retrograde researcher” and suggested that this study should “mark the beginning of the end of Mark Regnerus’s credibility with respectable news outlets.” These responses are a case study in how not to engage in constructive social science discourse.
To charge that the study is junk science disregards obvious facts about its publication. Social Science Research asked three experts to comment on the Regnerus study in the same issue of the journal. The experts counseled caution and noted some of the same interpretive limitations described above, but all three praised the study as an important contribution.
The facts are also at odds with the allegation of political bias. Regnerus took pains to assemble an ideologically diverse group of researchers to help in planning his study and supervising the data collection. No evidence indicates that any of the sources of his funding, which came in part from conservative organizations, played any role in the development or analysis of the NFSS. 
These facts about the NFSS design contrast with the allegations of a blogger-activist who claimed that the study was “designed so as to be guaranteed to make gay people look bad, through means plainly fraudulent and defamatory.” The blogger lodged an official complaint with the University of Texas, which triggered an automatic “scientific misconduct” inquiry into Regnerus’s work. On August 29, the university issued a press release exonerating Regnerus and closing the inquiry.
Regrettably, the denunciations and personal attacks were not focused solely on Mark Regnerus and his study. Social Science Research and its editor were subjected to similar attacks. A joint letter to the editor signed by numerous academics alleged that the journal did not apply its usual level of scrutiny to the paper, but the letter provided no evidence for this claim. The letter also criticized the choice of reviewers, acknowledging that they are “certainly well respected scholars” but complaining that they have never “published work that considers LGBT family or parenting issues.” Left unstated was what such a researcher would understand about the Regnerus study’s methodology that three well-respected family scholars would not.
To dispel the controversy fanned by the joint letter and accusations in the press and blogosphere, the editor of Social Science Research requested an internal audit to review the publication process for Regnerus’s piece. The Chronicle of Higher Education reviewed the audit and reported that it “did not find that the journal’s normal procedures had been disregarded, or that the Regnerus paper had been inappropriately expedited to publication, as some critics have charged.”
The exaggerated and extreme reactions to the paper, particularly those that call for exiling Mark Regnerus from polite society, fall far below the expectations of scientific discourse, not to mention the standards of civil debate in general. Rather than substantively engage the study, many critics have attempted to discredit legitimate research with baseless denunciations, unfounded insinuations of editorial impropriety, and personal attacks. Sober, fair-minded analysis is especially important when research has implication for an issue as politically charged as same-sex marriage.
The APA’s claim that no differences exist between children of same-sex parents and children with heterosexual parents has been used as an argument in favor of same-sex marriage. However, it is inappropriate to draw such a general policy conclusion from the studies on which the APA’s claim was based—studies with small or unrepresentative samples that cannot be generalized to the population at large.
By contrast, a June 2012 study by Mark Regnerus helps to shed more light on the issue by using the New Family Structures Study, a nationally representative random sample. Regnerus’s study found that adult children of parents who had a same-sex relationship report a number of negative outcomes compared with those who had two married biological parents. It also suggests that differences may exist between children whose parents had a same-sex relationship and those in non-intact heterosexual households.
These research findings have met fierce and frequently uncivil opposition, presumably because of their implications regarding the debate over the definition of marriage. In light of the Regnerus study, existing social science data do not support a popular argument for same-sex marriage—the “no differences” claim. The subsequent slurs against the study and the ad hominem attacks on its author are wholly inappropriate in scientific discourse.
The new information provided by the Regnerus study should enhance—not preempt—debate about the important policy questions related to the institution of marriage. Both sides of the debate should welcome Regnerus’s research as a careful, data-driven contribution to an issue of such magnitude as the future of marriage.
—Jason Richwine, PhD, is Senior Policy Analyst for Empirical Studies in the Domestic Policy Studies Department and Jennifer A. Marshall is Director of Domestic Policy Studies at The Heritage Foundation.