In research and policy settings, children in unique distress with same-sex parents are not supposed to exist. Most studies have reported “no differences” in well-being, most often using psychometric measures of depression or anxiety, supporting a lapse in policy attention to the potential needs of such children. Uniformly benign findings for this population have recently been challenged, however, by several original research efforts [1–3], the rediscovery of older studies [4, 5], and the reanalysis of studies long thought to support “no differences” .
The sparse and gendered nature of the same-sex parent population largely restricts research in this area to the examination of small samples of lesbian parents. Unfortunately, this difficulty has prompted an almost universal dependence on convenience samples [7, 8] recruited, with knowledge of study goals, from internet surveys, “LGBT events, bookstore and newspaper advertisements, word of mouth, networking, and youth groups” . Reanalyses have confirmed, not surprisingly, the presence in such samples of strong ascertainment bias, social desirability bias, and/or positive reporting bias [9–11]. In most studies, lack of statistical significance using simple bivariate tests in such samples is then erroneously interpreted as strong evidence of “no differences” in the population, even when difference in estimates or effect sizes are substantively large and even though the sample is not representative .
In fact, only four of the several dozen studies alleging “no differences” have examined a representative sample. The largest and most recent of these, Rosenfeld’s analysis of 3,174 same-sex parented children on the US Census, is discussed in Section 5. The other three are related studies based on a single sample, a group of 44 adolescents with lesbian parents captured on over 20,000 population-representative cases of the initial wave of the National Longitudinal Survey of Adolescent Health (“Add Health”) [13–15]. Sullins, however, recently found that most (27 of the 44) adolescents in this sample allegedly with same-sex parents were actually living with opposite-sex parents including, for most of them, their biological father as well as their mother. After removing the mixed cases, the remaining sample members fared significantly worse on psychometric measures of anxiety and autonomy than did their adolescent counterparts with opposite-sex parents, albeit comprising only 17 cases . Other studies employing large representative samples have also found higher depressive symptoms, indicated by the Center for Epidemiological Studies Depression Scale (CES-D ), among younger same-sex parented children  and adults who report having had a same-sex-related parent at some point during childhood . The design and methodology of Regnerus’ study were the subject of a brief but vigorous debate [19–21], which turned largely on definitional issues .
No study has yet explored the connection, if any, between late onset distress and precipitating conditions in children in this population, and no research reporting “no differences” has yet investigated parental child abuse or adult onset difficulties . The present study aims to amend these gaps in the research. It improves on the sample limitations of prior studies by employing data that are both representative and longitudinal, following the corrected Add Health sample of adolescents with lesbian parents, the most well-regarded small sample used in this field to date, through Wave IV, thirteen years after the initial interview at age 15 (on average). It improves on prior methods by the use of standard psychometric scales, to the extent possible, and the estimation of relative risk by logistic regression models with appropriate survey weighting. As the first study to examine children raised by same-sex parents into early adulthood, this exploratory study aims to contribute new information for understanding of the effects of same-sex parenting through the life-course transition into early adulthood.
The analysis followed a grounded theory approach, first identifying the presence or absence of pertinent differences by family type and then developing and testing grounded hypotheses, drawing both from the observed bivariate characteristics of the data and prior research where applicable. For clarity the research presentation will also follow this order, with the formulation of hypotheses presented following initial bivariate results.