First Look at Mark Regnerus’s Study on Children of Parents In Same-Sex Relationships
June 10th, 2012
A new study slated for the July issue of the journal Social Science Research claims to show significant differences in adverse outcomes among children raised by gay and lesbian parents when compared to children raised by both biological parents in a heterosexual-headed household. The study is not yet online but I have been provided an advance copy. [Update: it is now available for purchase.] It is by Mark Regnerus of the University of Texas’s Department of Sociology and Population Research Center, and is titled, “How Different are the Adult Children of Parents Who Have Same-Sex Relationships? Findings from the New Family Structures Study.”
This study finds “numerous, consistent differences, especially between children of women who have had a lesbian relationship and those with still-married (heterosexual) biological parents.” The results of this study would up-end some thirty years of established scientific research which showed that gay and lesbian parents are, on the whole, just as good as their straight counterparts. It would, at least, if the study’s methodology were designed to prove that point. But as is the case with all studies, the conclusions drawn by this study are only as good as the methodologies used to inform them.
The Study’s Sample
On that score, there is one significant strength to this study which makes it stand out. Unlike prior studies, the New Family Structures Study (NFSS) is based on a national probability sampled population. This is the gold standard for all social science studies, and it’s extremely rare for a study to achieve that mark. As far as I am aware, all of the studies to date of gay and lesbian parenting use non-representative convenience samples. National probability samples, unlike convenience samples, are important because they alone can be generalized to the broader populations, to the extent that key characteristics in the design of the probability sample (demographics, etc.) match those of the general population. Convenience samples can’t do that. (For more information on convenience samples versus national probability samples, click here.)
So why don’t the other studies use national probability samples? Believe me, every researcher would much rather work with national probability samples instead of convenience samples. But virtually nobody can afford the huge cost of putting such a study together. It is a massive undertaking, and the cost of creating such a data set is just too prohibitive. Regnerus however has overcome this limitation (PDF: 74KB/12 pages) with a generous $695,000 grant from the Witherspoon Institute and a supplemental $90,000 grant from the Bradley Foundation. With more than three quarters of a million dollars, he has the kind of funding that other researchers can only dream of.
With those vast sums in hand, Regnerus contracted with Knowledge Networks, a large research firm which has provided access to its broad, general-population probability sample to researchers for more than 350 working papers, conference presentation, published articles and books. The probability sample supplied by Knowledge Networks for this study was based on adults between the age of 18 and 39 who were asked the following questions (page 756):
“From when you were born until age 18 (or until you left home to be on your own), did either of your parents ever have a romantic relationship with someone of the same sex?” Response choices were “Yes, my mother had a romantic relationship with another woman,” “Yes, my father had a romantic relationship with another man,” or “no.” (Respondents were also able to select both of the first two choices.) If they selected either of the first two, they were asked about whether they had ever lived with that parent while they were in a same-sex romantic relationship. The NFSS completed full surveys with 2988 Americans between the ages of 18 and 39.
Because the number of adult children of gay or lesbian parents is so small, additional participants were recruited who could answer the first question in the affirmative using the same random methodologies used to generate the broader sample. This was done to increase the statistical power of the smaller sample while preserving the random nature of that subsample. This is a legitimate practice for examining very small populations. But for that reason, comparing the total of children raised by a parent who had had a same-sex relationship to the overall sample size would not tell us how much of the general population is being raised by such a parent. And while that was not a primary question to be answered by this study, Regnerus notes that the original raw sample showed that 1.7% of all Americans between the ages of 18 and 39 report that their father or mother has had a same-sex relationship. This is in line with several other studies on same-sex households with children.
Problem #1: Identifying a Same-Sex Relationship Doesn’t Tell Us Anything About the Nature of the Relationship.
But this is where we run into the first problem with this study. Identifying a parent who has had a same-sex relationship is not the same as identifying a parent who is gay, lesbian or bisexual in a functional relationship. Regnerus acknowledges this problem so quickly that it’s likely to be overlooked (p 578):
In the results section, for maximal ease, I often make use of the acronyms IBF (child of a still-intact biological family), LM (child of a lesbian mother), and GF (child of a gay father). It is, however, very possible that the same-sex romantic relationships about which the respondents report were not framed by those respondents as indicating their own (or their parent’s own) understanding of their parent as gay or lesbian or bisexual in sexual orientation. Indeed, this is more a study of the children of parents who have had (and in some cases, are still in) same-sex relationships than it is one of children whose parents have self-identified or are ‘‘out’’ as gay or lesbian or bisexual. The particular parental relationships the respondents were queried about are, however, gay or lesbian in content. [Emphasis in the original.]
What criteria does he use to judge whether those relationships are “gay or lesbian in content”? We don’t know. I can’t find any indication that the duration of the relationship played any factor in its identification. The question was very simple: “Did either of your parents ever have a romantic relationship with someone of the same sex?” Did a short affair count? What about a relationship that only lasted two weeks? Four months? Social conservatives who reduce gay and lesbian relationships to behaviors only would say yes to all of those, many eagerly so. Actual gay and lesbian couples in committed, long-term relationships naturally scoff at that, and rightfully so.
But already you can see where this is headed, can’t you? We will be asked to accept as legitimate the comparison of children raised by parents in those less stable and unenduring situations with stable, longtime married heterosexual parents without knowing the answers to those question. So already, Regnerus puts that comparison in starkly unequal footing.
Sorting the Sample into Subpopulations
That population was then sorted into the following categories (p 757-758):
1. IBF: Lived in intact biological family (with mother and father) from 0 to 18, and parents are still married at present (N = 919).
2. LM: R reported R’s mother had a same-sex romantic (lesbian) relationship with a woman, regardless of any other household transitions (N = 163).
3. GF: R reported R’s father had a same-sex romantic (gay) relationship with a man, regardless of any other household transitions (N = 73).
4. Adopted: R was adopted by one or two strangers at birth or before age 2 (N = 101).
5. Divorced later or had joint custody: R reported living with biological mother and father from birth to age 18, but parents are not married at present (N = 116).
6. Stepfamily: Biological parents were either never married or else divorced, and R’s primary custodial parent was married to someone else before R turned 18 (N = 394).
7. Single parent: Biological parents were either never married or else divorced, and R’s primary custodial parent did not marry (or remarry) before R turned 18 (N = 816).
8. All others: Includes all other family structure/event combinations, such as respondents with a deceased parent (N = 406).
Problem #2: Arbitrary Decisions in Dealing with Overlaps Between Categories Make the LM and GF Categories Heterogeneous While the Other Categories Remain Relatively Homogeneous.
Regnerus discusses how he dealt with the fact that these categories are not mutually exclusive (page 758):
That is, a small minority of respondents might fit more than one group. I have, however, forced their mutual exclusivity here for analytic purposes. For example, a respondent whose mother had a same-sex relationship might also qualify in Group 5 or Group 7, but in this case my analytical interest is in maximizing the sample size of Groups 2 and 3 so the respondent would be placed in Group 2 (LMs). Since Group 3 (GFs) is the smallest and most difficult to locate randomly in the population, its composition trumped that of others, even LMs. (There were 12 cases of respondents who reported both a mother and a father having a same-sex relationship; all are analyzed here as GFs, after ancillary analyses revealed comparable exposure to both their mother and father).
This quoted paragraph packs quite a wallop. Regnerus describes “small minorities” which overlap different categories, but he doesn’t characterize them except for one example — the twelve respondents who reported both a mother and a father having a same-sex relationship. He decided to lump them into the Gay Fathers (GF) category rather than the Lesbian Mothers (LM) category. That alone boosts the GF numbers by one-fifth. That’s a “small minority” in an election, but in a statistical study it’s not so small.
Regnerus’s decision to arbitrarily force his sample into non-overlapping categories results in a method is based solely on a desire to increase the size of the smaller group, a goal which has nothing to do with the study’s larger goal of comparing children of gay and lesbian parents to those of intact biological families. In fact, he makes choices which, by their very nature, run explicitly counter to that goal. With each transfer of a subject from the Divorced, Step Families, Single Parent or “All Others” categories into LM or GF, the less those LM and GF groups are designed to look like intact biological families. Especially if you consider some of the possibilities that might exist in the “All Others” category (page 757):
Fifty-eight (58) percent of those whose biological mothers had a same-sex relationship also reported that their biological mother exited the respondent’s household at some point during their youth, and just under 14% of them reported spending time in the foster care system, indicating greater-than-average household instability. [Emphasis mine.]
I understand Regnerus’s problem with having a sample of lesbian mothers and gay fathers that is too small to make statistical comparisons with the control group. But the proper solution to that problem would have been to go back to Knowledge Networks for another round of sample recruitment. Then perhaps he could have compared children of gay and lesbian parents who had been brought up in a stable environment with children of heterosexual parents in a stable environment. He could have also compared children of gay and lesbian parents in unstable environments with their heterosexual counterparts.
But instead, he undertakes a manipulation which I believe represents the fatal flaw of this study. If one wanted to intentionally create Lesbian Mothers and Gay Fathers groups which were least likely to look like an intact biological family, I can’t imagine a better way to do so than to take the steps Regnerus has taken here. He enlarged his LM and GF groups by lumping together a mishmash of overlapping characteristics into two messy samples. The other six categories are relatively homogenous for straight people, but Regnerus’s enlargement of LM and GF groups makes them deliberately heterogeneous. And now having done this, he’s about to compare two deliberately heterogeneous categories (LM and GF) to a deliberately homogeneous category. Well of course there will be differences once you do that!
Problem #3: It Doesn’t Study Children Who Grew Up In Gay- or Lesbian-Led Households.
Marriage equality opponents will trumpet this study as proof that children raised by loving, committed, married same-sex couples will have more problems than those who are raised by both biological parents in a heterosexual household. But the samples that Regnerus put together to compare to the ideal heterosexual household cannot make such a claim because it says almost nothing about committed same-sex couples who, together, decide to become parents via adoption, in-vitro fertilization or surrogacy (page 756):
While gay and lesbian Americans typically become parents today in four ways — through one partner’s previous participation in a heterosexual union, through adoption, in-vitro fertilization, or by a surrogate—the NFSS is more likely to be comprised of respondents from the first two of these arrangements than from the last two. Today’s children of gay men and lesbian women are more apt to be ‘‘planned’’ (that is, by using adoption, IVF, or surrogacy) than as little as 15–20 years ago, when such children were more typically the products of heterosexual unions. The youngest NFSS respondents turned18 in 2011, while the oldest did so in 1990. Given that unintended pregnancy is impossible among gay men and a rarity among lesbian couples, it stands to reason that gay and lesbian parents today are far more selective about parenting than the heterosexual population, among whom unintended pregnancies remain very common, around 50% of total (Finer and Henshaw, 2006). The share of all same-sex parenting arrangements that is planned, however, remains unknown. Although the NFSS did not directly ask those respondents whose parent has had a same-sex romantic relationship about the manner of their own birth, a failed heterosexual union is clearly the modal method: just under half of such respondents reported that their biological parents were once married. This distinguishes the NFSS from numerous studies that have been entirely concerned with ‘‘planned’’ gay and lesbian families, like the NLLFS.
What Regnerus sees as a strength, I see as a gross weakness in the context of current debates over same-sex marriage. Those debates center on whether committed, loving same-sex relationships are worthy of recognition, and one component of that debate has been what effects, if any, loving, same-sex relationships have on children. But to have a legitimate comparison, you’d have to look at children whose experience in that same-sex family was, for all intents and purposes, in place for their entire or almost entire childhoods. After all, that is the experience of children from intact biological heterosexual families.
But in this study, only 57% said they had lived with their mother and her partner for at least four months before the age of 18, and only 23% reported living with their father and his partner for the same length of time. Only 23% of LM children and 2% of GF children reported living with their parents and their parents’ same-sex partners for three years or more. And when looking at the outcomes of those children, we are being led to believe that those outcomes are in some way related to the short amounts of time that those children spent with their gay or lesbian parents while in a same-sex relationship, and not the fifteen-plus years the vast majority of them spent outside of that dynamic. The illogic behind this comparison is mind-boggling.
So, we’ve been talking about the results of the study — that children of parents who have had a same-sex relationship have greater struggles than those who grow up in intact biological families — without actually seeing those results. It’s time now to look at them. The first table of results, Table 2, shows the average responses given by these now-adult children to several yes/no questions, including:
- Are you currently married? Cohabiting?
- Did your family receive welfare growing up?
- Are you currently employed?
- Have you recently thought about suicide?
- Do you identify entirely as heterosexual?
- Have y0u had an affair while married/cohabiting?
- Have you ever been touched sexually by a parent/adult?
- Have you ever been force to have sex against your will?
Anti-gay extremists are likely to latch onto the last two questions for evidence that children of gay and lesbian parents are particularly vulnerable to sexual abuse at the hands of those parents. But notice the open-ended nature of the questions: they don’t ask who did the abuse, but only whether the abuse occurred. As Regnerus explains (page 763):
It is entirely plausible, however, that sexual victimization could have been at the hands of the LM respondents’ biological father, prompting the mother to leave the union and—at some point in the future—commence a same-sex relationship. Ancillary (unweighted) analyses of the NFSS, which asked respondents how old they were when the first incident occurred (and can be compared to the household structure calendar, which documents who lived in their household each year up until age 18) reveal this possibility, up to a point: 33% of those LM respondents who said they had been sexually victimized by a parent or adult caregiver reported that they were also living with their biological father in the year that the first incident occurred. Another 29% of victimized LMs reported never having lived with their biological father at all. Just under 34% of LM respondents who said they had at some point lived with their mother’s same-sex partner reported a first-time incident at an age that was equal to or higher than when they first lived with their mother’s partner. Approximately 13% of victimized LMs reported living with a foster parent the year when the first incident occurred. In other words, there is no obvious trend to the timing of first victimization and when the respondent may have lived with their biological father or their mother’s same-sex partner, nor are we suggesting by whom the respondent was most likely victimized.
Wanna bet the Family “Research” Council ignores that?
Moving on to Table 3, we see the mean scores to several continuous outcome variables, which include:
- Educational attainment.
- Family of origin safety/security
- Closeness to biological mother/father
- Physical health, overall happiness, depression
- Attachment, impulsivity measures
- Current relationship quality measures
And finally, Table 4 shows the mean scores to several event-count outcome variables, which include:
- Frequency of marijuana/alcohol/tobacco use
- Frequency of watching TV
- Frequency of arrests, pleading guilty to non-minor offenses
- Number of female/male sex partners
A cursory look at all of this data reveals what we already know: children who have experienced divorce, single-parenthood, step-parenthood, and other family structures other than a long-term stable family experience all kinds of problems, regardless of whether their parents are gay or straight. And you know what? Regnerus himself says so, but with very different words:
(Page 763): Although my attention has been primarily directed at the inter-group differences between IBFs, LMs, and GFs, it is worth noting that LMs are hardly alone in displaying numerous differences with IBFs. Respondents who lived in stepfamilies or single-parent families displayed nine simple differences in Table 2. Besides GFs, adopted respondents displayed the fewest simple differences (three).
(Page 763-4): As in Table 2, respondents who reported living in stepfamilies or in single-parent households also exhibit numerous simple statistical differences from IBFs—on nine and 10 out of 14 outcomes, respectively—most of which remain significant in the regression models. On only four of 14 outcomes do adopted respondents appear distinctive (three of which remain significant after introducing controls).
By the time he gets to discussing Table 4, Regnerus is no longer interested in making the broader observation, saying instead (page 764):
Although I have paid much less attention to most of the other groups whose estimates also appear in Tables 2–4, it is worth noting how seldom the estimates of young-adult children who were adopted by strangers (before age 2) differ statistically from the children of still-intact biological families. They display the fewest simple significant differences– seven — across the 40 outcomes evaluated here. Given that such adoptions are typically the result of considerable self-selection, it should not surprise that they display fewer differences with IBFs.
It’s worth noting his comment about adoptive parents being intentional parents. By his own admission, that is an emerging model of parenthood for gay and lesbian parents in committed long-term relationships going forward. And that point leads us directly to…
Problem #4: This Study Makes The Wrong Comparison.
When you look at the data, the study’s real findings become obvious. Children of parents who have had a same-sex relationship — a group that includes very large numbers of children of divorced parents, single parents, adopted parents, step-parents and “other” family structures — have developmental outcomes which are remarkably similar to children of divorced, single, adopted, step-, and “other” family structures overall when compared to intact, non-adoptive heterosexual families. Regnerus designed his study to show this result by constructing samples which mimicked these characteristics. By constructing his LM and GF samples the way he did, the only legitimate comparison he could make would be to children of divorced, single, adopted, step-, and “other” family structures. But that’s not the comparison he made. He focused the study on making the wrong comparison, and then concluded that children of gay and lesbian parents have more negative outcomes than children of straight parents in intact households.
But his conclusion is as illegitimate as his comparison, and his comparison is as illegitimate as his samples. Instead of making an apples to apples comparison, he compared apples to elephants and concluded that there was a difference. To which anyone with a minimal understanding of design of experiments would answer, Duh!
But when you make the right comparison — compare children of divorced gay parents to children of divorced straight parents, compare children of single gay parents to children of single straight parents, compare children of long-term committed gay parents to children of long-term straight parents — there is nothing in this study to suggest that children of gay parents are at any disadvantage whatsoever to children of straight parents. And when you make the right comparison, the study goes much further toward confirming that conclusion than the one Regnerus arrives at.
The Study’s Caveats
Alongside the study’s explosive claims, Regnerus does provide several important caveats. I want to note them and make sure they are publicly available, since anti-gay groups are likely to make claims about this study that Regnerus doesn’t make.
(Page 755): There are several things the NFSS is not. The NFSS is not a longitudinal study, and therefore cannot attempt to broach questions of causation. It is a cross-sectional study, and collected data from respondents at only one point in time, when they were between the ages of 18 and 39. It does not evaluate the offspring of gay marriages, since the vast majority of its respondents came of age prior to the legalization of gay marriage in several states. This study cannot answer political questions about same-sex relationships and their legal legitimacy.
(Page 766): As scholars of same-sex parenting aptly note, same-sex couples have and will continue to raise children. American courts are finding arguments against gay marriage decreasingly persuasive (Rosenfeld, 2007). This study is intended to neither undermine nor affirm any legal rights concerning such.
(Page 766): Although the findings reported herein may be explicable in part by a variety of forces uniquely problematic for child development in lesbian and gay families—including a lack of social support for parents, stress exposure resulting from persistent stigma, and modest or absent legal security for their parental and romantic relationship statuses—the empirical claim that no notable differences exist must go.
On that last point, let me say that I would have to agree. To claim that there are no notable differences in childhood outcomes due to family stability or structure would be ludicrous. And I don’t think it’s possible to legitimately claim that there are no differences between gay- and lesbian-led families and their straight counterparts. There are bound to be differences. But thirty years of research has failed to find major adverse differences where appropriate comparisons were made. The data in this study appears to reinforce that fact rather than refute it — when appropriate comparisons are made. But with this nationally representative sample, we’re still waiting for those appropriate comparisons.
A Final Note
I made a brief mention of the huge amount of money that was spent on this study, a sum that comes to $785,000. The lion’s share came from the Witherspoon Institute, a think tank in Princeton, New Jersey (and not affiliated with Princeton University). Members of the Institute include Robert P. George, who drafted the Manhattan Declaration and whose recent paper in The Harvard Journal of Law and Public Policy on same-sex marriage was critiqued at length by BTB’s Rob Tisinai. The Withersoon Institute reportedly has close associations with such organizations as the National Organization for Marriage, the Family Research Council, and the secretive Catholic order Opus Dei. George also sits on the board of directors for the Bradley Foundation, which also provided funds for this study. The Bradley Foundation is considered one of the country’s largest and most influential right-wing foundations, although its contribution to this study is “only” $90,000.
I mention this because it is important, but I don’t want to over play its importance. All sorts of studies are funded by all sorts of institutions which support a variety of causes. Those sources can come from conservative, anti-gay organizations, or (as is the case with many studies which are favorable to LGBT issues) they can come from pro-gay sources such as the Williams Institute or other organizations. The source of funding can indicate a potential conflict of interest, but the true value of a study rests on the methodology of the study itself. If the methodology is sound, then the study’s conclusions are sound regardless of where the money comes from or who’s doing it. But where the methodology fails, the broken link affects the entire chain. While there are many grounds in which to attack this study, the only legitimate way to critique it is to examine the methodology, as I have done here.
A Final, Final Note
I only received this study yesterday morning, which means that my examination was the result of only one day’s work on a Saturday. So consider this examination a preliminary one. I look forward to seeing what you learn in the comments or via email for a possible future update.