ArticlePDF Available

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

Authors:

Figures

Content may be subject to copyright.
_____________________________________________________________________________________________________
*Corresponding author: E-mail: sullins@cua.edu;
British Journal of Education, Society &
Behavioural Science
11(2): 1-22, 2015, Article no.BJESBS.19337
ISSN: 2278-0998
SCIENCEDOMAIN international
www.sciencedomain.org
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
D. Paul Sullins
1*
1
Department of Sociology, The Catholic University of America, USA.
Author’s contribution
The sole author designed, analyzed and interpreted and prepared the manuscript.
Article Information
DOI: 10.9734/BJESBS/2015/19337
Editor(s):
(1) Daniel Yaw Fiaveh, Department of Sociology, University of Ghana, Legon, Ghana.
(2)
Stan Weeber, Professor of Sociology, McNeese State University in Lake Charles, Louisiana,
USA.
Reviewers:
(1) Anonymous, University of Maiduguri, Nigeria.
(2)
Paulo Verlaine Borges e Azevêdo, University of Goiás, Brazil.
(3)
Anonymous, University of Sri Jaywardenepura, Sri Lanka.
Complete Peer review History: http://sciencedomain.org/review-history/10443
Received 4
th
June 2015
Accepted 21
st
July 2015
Published 8
th
August 2015
ABSTRACT
Aims:
To critique, replicate and re-analyze Wainright and Patterson’s three studies of adolescents
with same-sex parents, which conclude, based on representative population data, that such
children suffer no disadvantages.
Methodology: After replicating Wainright and Patterson’s sample and analyses using the National
Longitudinal Survey of Adolescent Health, Wave I, (n = 20,745), re-examination of the same-sex
parent sample finds that 27 of the 44 cases are misidentified heterosexual parents; they did not
adjust for survey design and clustering; and ignored 99 percent of the baseline by using a small
matched sample for comparison. Outcomes are re-analyzed after correcting these problems, using
OLS, logistic regression and Firth (bias-adjusted) regression models.
Results: The adolescents with same-sex parents experience significantly lower autonomy and
higher anxiety, but also better school performance, than do adolescents with opposite-sex parents.
Comparing unmarried to (self-described) married same-sex parents, above-average child
Original Research Article
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
2
depressive symptoms rises from 50% to 88%; daily fearfulness or crying rises from 5% to 32%;
grade point average declines from 3.6 to 3.4; and child sex abuse by parent rises from zero to 38%.
The longer a child has been with same-sex parents, the greater the harm.
Conclusion: Children with same-sex parents experience significant disadvantages, but also some
advantages, compared to those with man-woman parents. Although opposite-sex marriage is
associated with improved outcomes on a wide range of child well-being measures, same-sex
marriage is associated with lower outcomes. Further work is needed to determine the relative
influences of instability, duration, and marriage to these findings.
Keywords: National Longitudinal Survey of Adolescent Health; same-sex parents; child well-being;
same-sex marriage.
1. INTRODUCTION
Since the 1970s a rapidly-growing body of
empirical studies has compared homosexual and
heterosexual relationships and parenting
outcomes, concluding almost without exception
that relationship quality and human flourishing in
homosexual relationships is equivalent to that in
heterosexual ones and that children raised by
homosexuals do not suffer adverse harm (the no
differences thesis). Almost all such results have
been based on small, non-random samples,
usually consisting of participants recruited for
convenience who are aware of the purpose of
the study, and for this reason have failed to be
convincing.
This problem has been noted repeatedly by
scholars adopting different widely different
opinions on the underlying question of same-sex
parenting. For example, Wendy Manning and
colleagues, reviewing the literature for a court
brief supporting same-sex marriage, counted
studies of only four large random samples,
noting: “Convenience samples are more common
.... Relying on convenience samples means that
the same-sex parents in these studies are not
representative of all same-sex parents and
represent only those who were targeted and
agreed to participate, ….” [1]. Likewise Michael
Rosenfeld, in a study finding no differences in
school outcomes with same-sex parents,
observed: “As the critics have noted,
convenience samples dominated this literature in
the past” [2]. Douglas Allen, in a rebuttal of
Rosenfeld’s study that found lower graduation
rates for children with same-sex parents, agreed:
“Although a proper probability sample is a
necessary condition for making any claim about
an unknown population, within the same-sex
parenting literature researchers have studied
only those community members who are
convenient to study” [3].
As all three authors just cited acknowledge, a
notable exception to the use of convenience
samples has been three related studies that
made use of data from the National Longitudinal
Survey of Adolescent Health (“Add Health”). The
first study, published in 2004 by Wainright,
Russell and Patterson (hereafter “WRP 2004”),
explored the connections between psychosocial
well-being, school performance, and romantic
relationships in the two family types [4].
Wainright and Patterson (hereafter “WP”)
followed up with a brief report in 2006 looking at
delinquency, victimization and substance abuse
[5], and a 2008 study of peer relations [6]. A
2009 review by Patterson summarizes all three
studies [7].
By most accounts, including Rosenfeld’s [2],
these studies are the only ones prior to
Rosenfeld’s 2010 study to employ a
representative population sample with sufficient
statistical power to discern differences, if they
existed, for children with same-sex parents (but
see [8]). All three studies examined the same
sample, a group of 44 adolescents with lesbian
mothers on the initial wave of the National
Longitudinal Survey of Adolescent Health, which
surveyed over 20,000 adolescents in 1995. The
design features of the analysis are similar in all
three studies, comparing the adolescents with
lesbian mothers with a matched group of
adolescents with heterosexual parents; the main
analytic differences (as distinct from the
theoretical questions involved) have to do with
the examination of different outcome variables in
each study. The studies refer to the two groups
of same-sex parents and opposite-sex parents
as “family types”, a wording I will also adopt for
simplicity in the present study.
All three WP studies concluded that, on the
variables examined in the study, “adolescents
living with same-sex parents did not differ from
that of adolescents living with opposite-sex
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
3
parents” [4] in any way that would disadvantage
the former. With respect to this conclusion the
authors are aware that their results “add
significantly to those from earlier studies, which
were most often smaller in their size, less
representative in their sampling, and less
comprehensive in their assessment of
adolescent outcomes” [4]. Indeed, these three
studies present some of the strongest evidence
in support of the no differences thesis, and for
that reason are often cited prominently in
subsequent research and in legislative and
judicial policy settings.
Subsequent studies of other representative data,
however, have failed to confirm most of WP’s
conclusions. In a representative sample of 2,988
adults in 2012, Regnerus found significantly less
positive outcomes on a wide range of
psychosocial, relational and functional measures
for a group of 248 adults whose parent or
parents had ever been in a homosexual
relationship [9]. Sullins, examining over 200,000
cases from the National Health Interview Survey
that included 512 children with same-sex
parents, found that emotional problems, including
anxiety, and other indicators of psychosocial
distress, were more than twice as prevalent
among children with same-sex parents [10]. The
only conclusion of WP that may possibly have
been replicated is Rosenfeld’s 2010 claim, based
on a large sample from the U.S. Census, that
children with same-sex parents progressed
normally through school [2]. However, Allen
failed to replicate Rosenfeld’s finding using the
Canadian census [3] and has disputed
Rosenfeld’s analysis [11].
To address this difficulty, the current study
attempts to critically evaluate and replicate WP’s
2004 conclusions, and if feasible to re-analyze
their original data, in order to confirm or counter
their findings with a greater degree of confidence
than has previously been the case.
2. DATA AND METHODS
This research uses data from Add Health, a
program project directed by Kathleen Mullan
Harris and designed by J. Richard Udry, Peter S.
Bearman, and Kathleen Mullan Harris at the
University of North Carolina at Chapel Hill, and
funded by grant P01-HD31921 from the Eunice
Kennedy Shriver National Institute of Child
Health and Human Development, with
cooperative funding from 23 other federal
agencies and foundations. Special
acknowledgment is due Ronald R. Rindfuss and
Barbara Entwisle for assistance in the original
design. Information on how to obtain the Add
Health data files is available on the Add Health
website (http://www.cpc.unc.edu/addhealth). No
direct support was received from grant P01-
HD31921 for this analysis. The author’s
management and use of the data has been
reviewed and approved by the Institutional
Review Board of the Catholic University of
America.
Add Health, also known as the National
Longitudinal Survey of Adolescent Health, has
followed a large random sample of American
adolescents for twenty years. Wave I was
administered in 1995 through in-school
interviews with over 90,000 American
adolescents aged 13-19 selected by means of a
stratified random sample of U.S. high schools.
Of these, 27,000 were selected for a more
extensive interview in their home and a separate
related interview with their mother. If the mother
was not available after separate attempts, the
father or another adult in the household was
interviewed. The in-home interview sample
consisted of a core sample selected randomly
using a complex multi-stage sampling process
that was stratified by region, other strata, and
geographic areas known as probability sampling
units. A total of 12,105 core sample interviews
were augmented by an additional 8,640 cases
that reflect a series of oversamples and special
interest data groups, to comprise the full sample
of 20,745 cases. Through the application of post-
stratification weights that reflect known
characteristics of the adolescent population at
that time, the sample is rendered representative
of the U.S. adolescent population with a high
degree of precision.
The current study replicates the sample and
mean comparisons of WRP 2004 using t-tests in
place of the original ANOVA, and employs
logistic regression models to assess differences
between family types. All analyses were
performed with Stata 13 statistical software,
incorporating the design features of the survey
following guidelines for analyzing Add Health
data published by the Carolina Population Center,
University of North Carolina [12].
3. VARIABLES IN THE ANALYSIS
The outcome variables examined by WRP were
replicated, as far as possible, from the
description provided in their study. Depressive
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
4
symptoms were measured by a 19-item version
of the Center for Epidemiologic Studies’
Depression Scale (CES-D) which was
administered in the in-home interview [13]. The
items in the scale name a list of symptoms such
as feeling sad, lonely, tired or bothered about
things. The response range for each item is from
0 (never or rarely) to 3 (most of the time or all of
the time); the range of the 19-item scale is from 0
to 57.
WRP 2004 reported that they measured
adolescent anxiety “with a seven-item scale from
the In-Home Interview that included questions
about the frequency of symptoms such as feeling
moody or having trouble relaxing”. These two
items are part of a six-item series (not seven) on
anxiety, which asks about both physical
conditions such as sleeplessness or poor
appetite as well as more direct indicators of
emotional distress such as moodiness,
fearfulness or frequent crying. The in-home
interview asks how often the respondent has
experienced each condition in the past twelve
months, with possible responses of “never, “just
a few times”, “about once a week”, “almost every
day”, and “every day”, coded from 0 to 4. The
present study uses these six items to form a
scale as close as possible to that used by WRP,
and in any event to effectively measure anxiety.
The item “Daily fearfulness/crying” in Table 4 is
derived from this scale, reporting the proportion
who responded “every day” or “almost every day”
for the items “fearfulness” or “frequent crying”.
Although WRP reported that their anxiety scale
ranged from 0 to 28, and reported a
corresponding number in the tables, in the text
they reported a mean anxiety score based on a
scale from 0 to 4. To ensure comparability the
anxiety scores reported in Table 2 are also
standardized on a 0-4 scale.
WRP reported that they summed 6 items with a
response scale of 1 to 5 to produce a self-
esteem scale ranging from 6 to 30, but report a
mean value of 4.02 for the same-sex sample,
and for each item report the reverse of the scale
shown on the Add Health file. I took the mean of
the reverse-coded items as the best guess at
what they actually did. The results of this scale
are consistent with the numbers they report [4].
Grade point average was reported on a scale
from 0 to 4.0. For school connectedness and
neighborhood integration WRP report the reverse
of the true scoring scale; it appears that they
recoded the items, so I did as well.
The Add Health in-home interview asked female
adolescents, “Were you ever physically forced to
have sexual intercourse against your will?”
Males were asked, “Did you ever force someone
to have sexual intercourse against her will?”
About one in ten respondents (11.6%, 95% CI
10.5-12.7) overall reported forcing or being
forced to have sex. In Table 4, where this
variable is introduced, both opposite-sex
categories and same-sex unmarried contain both
male and female respondents, but only female
respondents reported forced sexual intercourse
in the same-sex marriages, almost all of which
involve lesbian partners.
4. ANALYSIS
The analysis proceeded in three steps. The first
step was a critical appraisal of the elements of
the WRP 2004 study with regard to the possibility
of identifying differences for adolescents with
same-sex parents. Two features of the sample of
same-sex parents defined by WRP obscured its
effectiveness for identifying differences for
children with same-sex parents: the sample
mistakenly included a majority of cases that are
actually heterosexual parent couples, and the
sample excluded male same-sex couples. After
correcting these issues, the second step involved
replicating the analyses of WRP 2004, as far as
possible, to examine the effect, if any, of
amending the sample defects on the study
outcomes. Third, the corrected sample was
employed to examine the effect of marriage on
child outcomes with same-sex parents.
4.1 Step One: Critical Appraisal
4.1.1 Miscoded heterosexual parents
WRP identified same-sex parents by comparing
the sex of the responding mother with the
reported sex of a partner with whom she reported
that she was married or living in a marriage-like
relationship. They explain their procedure:
We first identified families in which parents
reported being in a marriage or marriage-like
relationship with a person of the same sex. …
[Then] the consistency of parental reports
about gender and family relationships was
examined. To guard against the possibility
that some families may have been
misclassified because of coding errors, we
retained only cases in which parental reports
of gender and family relationship were
consistent (e.g., a parent reported being
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
5
female and described her relationship to the
target adolescent as ‘‘biological mother ’’).
The focal group of families identified through
this process consisted of 44 adolescents, 23
girls and 21 boys. Approximately 68% of the
adolescents identified themselves as
European American or White, and 31.8%
identified themselves as non-White or as
biracial. On average, the adolescents were
15.1 years of age (SD 5 1.5 years), with a
range from 12 to 18 years of age [4].
In a related table, they also report that 4.5% of
these adolescents were adopted.
Following these procedures, I also found 44
adolescent cases on the Add Health sample
whose female parent respondent reported being
in a marriage or marriage-like relationship with a
another woman. I found no inconsistent parental
reports of gender and family relationships. This
group of 44 cases consisted of 23 girl and 21
boys (52.3% female), was 68% white with an
average age of 15.1 years, and 4.5% were
adopted. Since these characteristics exactly
match those reported by WRP above, I
concluded that this group is the same lesbian
parent sample identified in their study.
In the Add Health in-home interviews, responding
adolescents were asked to identify the sex and
relationship to themselves of all members of the
household. WRP 2004 reported that they
explored another consistency check for the
same-sex partners, which “required that if an
adolescent reported living with his or her
biological mother, he or she reported no male
figure (e.g., biological father, stepfather) as
residing in the household.” Applying this criterion,
they identified 18 cases which clearly consisted
of adolescents living only with two adult parents
of the same sex. Remarkably, they rejected this
criterion, even though it incorporates effectively
the same safeguard against misclassification as
the similar check they report using on the
parental interview. Their justification for this is
that they believed that “application of the more
stringent criteria effectively eliminated from the
sample many adolescents from divorced families
in which one or both parents were currently
involved in same-sex relationships” [4] as well as
children in joint custody arrangements.
It is hard to know what they mean by this. The
Add Health interview only asked responding
adolescents about persons “who live in your
household” [14]. If the adolescent reported the
presence of a father or father figure in this series
of questions, this could not have been a father in
another household, as would be the case in a
joint custody situation. In fact, of the 44 sample
adolescents, half (22) [Author: This is not a
reference, but the number 22.] of them reported
that their biological father lived in the home. An
additional four identified one of the household
members as their step or adoptive father, and
one reported the presence of a foster father. In a
separate question that asked the adolescent
independently to confirm the sex of each person,
all 27 of these fathers were explicitly identified as
male.
In a series of questions about non-resident
biological fathers, only the 18 clear cases of
adolescents living with two female same-sex
parents indicated any knowledge of a non-
resident father. Three of the four adolescents
who identified an adoptive or foster father were
assumed to have a non-resident biological father,
but they reported they did not know anything
about him. It is quite clear, in other words, that
only among the 18 clear cases could there
possibly be anything like a joint custody
arrangement. Five parents among the 18 clear
cases, but only one among the additional 26
cases including by rejecting the criterion of
having two same-sex parents, indicated that he
or she was divorced. Thus it is not the case that
the more stringent criteria “eliminated from the
sample many adolescents from divorced families”
[4].
Clearly, the 27 families for which the child
reported the presence of a resident male father
cannot reasonably be considered lesbian parent
families. Probably they are miscoded opposite-
sex families. At the very least, it is fair to say that
the sex designation is inconsistent, and, on the
same principle that WRP already screened out
cases with inconsistent parental reports of sex,
these cases should also be discarded. Excluding
these cases leaves 17 cases that are clearly and
consistently identified as lesbian parent couples.
WRP report finding 18 cases in this group; it is
possible that they include the one household
where the adolescent identified a “foster father”.
WRP note that the group identified by this more
stringent criteria has “the advantage of including
only clear cases in which adolescents described
themselves as living only with two same-sex
adults, and in which parents described
themselves as unmarried and as involved in a
marriage or marriage-like relationship with a
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
6
person of the same sex. In short, these families
conformed in every particular to an idealized
image of lesbian mother families” [4].
Nonetheless, they rejected using this sample.
4.1.2 Other design difficulties
Three other elements of WRP’s study design
obscured possible differences for adolescents
with same-sex parents. First, WRP compared
boys and girls separately within each family type,
despite having already matching the two
comparison groups on sex. This analytical choice
responds to other interests in their study, but it
also reduces each of the already-small family
type groups by about half.
Second, and more seriously, instead of
comparing the children with same-sex parents
with the full remaining sample of approximately
20,000 children, WRP compared them to another
group of 44 children matched to the children with
same-sex parents on a number of demographic
characteristics. A matched comparison like this is
an acceptable way to control for differences in
age, sex, parent education and income, etc., but
in this case, since the groups are so small to
begin with, doing so renders it needlessly more
difficult to show differences between the groups.
Compared with matched samples, correcting for
demographic differences by the use of control
variables is much more common in social
science analysis, since it preserves the ability to
standardize the groups on relevant demographic
characteristics while retaining the statistical
power of the entire dataset. Instead of comparing
a small group with large standard errors to a
large group with correspondingly small standard
errors, WRP compared two small groups, both of
which have large standard errors. Essentially,
WRP ignored 99% of the baseline, nullifying the
power of the large Add Health sample.
Third, WRP 2004 made no use of Add Health’s
complex survey design or post-stratification
weights, as elements of their analysis make clear.
They reported, for example, that they created the
list of matched adolescents with opposite-sex
parents “by generating a list of adolescents from
the Add Health database who matched each
target adolescent on the following characteristics:
sex, age, ethnic background, adoption status
(identified through parent reports), learning
disability status, family income, and parent’s
educational attainment. The first matching
adolescent on each list was chosen as the
comparison adolescent for that target
adolescent.” Since each unweighted case
represents a large and variable number of
weighted cases, this kind of one-to-one matching
could only have been accomplished using
unweighted cases. It is difficult to determine what
effect this omission may have, if any, on the
ability to identify differences for the adolescents
with same-sex parents, but it is a consequential
error that undermines confidence in the
representativeness of the study.
The Add Health Core Sample, which is based on
a stratified random sample of U.S. high schools,
could arguably be taken as roughly
representative of the adolescent population
without weighting. WRP claim that their analysis
was based on the core sample, in which case the
lack of weighting might not be a problem, or
much of a problem, but this claim cannot be true:
of the 44 cases in their sample of children
allegedly with same-sex parents, only 29 are in
the Core Sample. The additional 15 cases, and
thus the full sample, are made representative of
the population only by the application of post-
stratification weights.
Add Health’s Guidelines for Analyzing Add
Health Data advise: “To obtain unbiased
estimates, it is important to account for the
sampling design by using analytical methods
designed to handle clustered data collected from
respondents with unequal probability of selection”
[12]. In a section on common errors when using
Add Health, the first error listed is “Ignoring
clustering and unequal probability of selection
when analyzing the Add Health data” (boldfaced
in original) [12]. Since they ignored clustering,
WP’s findings cannot statistically represent the
population of same-sex parents, even if the
sample were accurate. They may, of course, be
suggestive in a general way. At best, these three
studies present findings from another
unrepresentative small group of same-sex
parents, such as are almost universal in this area
of research.
4.2 Step Two: Replication
4.2.1 Replication with the original sample
(and alternative partitions)
WRP also found 6 male same-sex partners in the
Add Health sample, but report that they excluded
them from their sample in order to focus more
clearly on lesbian parents, after preliminary
analyses that included the 6 male same-sex
partners produced results that “were very nearly
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
7
identical to those including only [the 44] families
headed by female same-sex couples.” Likewise,
they reported that they “completed all the
analyses” with the smaller group of 18 clear
cases of same-sex parents and the “results were
essentially identical” to those of the larger group
of 44 cases.
These claims may be a bit overstated, but they
are essentially accurate. Table 1 replicates
WRP’s analysis, as closely as possible, showing
results for their full sample (44 cases) and the
alternative sample groups or partitions
discussed: The “clear cases” of lesbian couple
parents (17 cases), the remainder of WRP’s 44
case same-sex parent sample, almost all of
whom are actually heterosexual parents (27
cases), and male same-sex parents (6 cases).
The table replicates WRP’s method of analysis,
comparing group mean values for each of the
outcome variables of interest. Only individual
outcomes are assessed, ignoring WRP’s
multivariate analyses. Rather than the ANOVA
tests reported by WRP, the table reports the
more commonly-used t-tests; t-tests and ANOVA
produce statistically identical decision results for
mean comparisons. Consistent with
recommended standards and other research on
the small population of same-sex parents, the
table also identifies group differences at the more
relaxed .10 level of significance, as well as the
conventional .05 level.
Columns A and B of Table 1 are derived from
WRP 2004, Table 2, with results interpolated by
sex, comparing the matched sample of 44
opposite-sex parents with their full sample of 44
(alleged) same-sex parents. WRP did not show
the p-values, but reported that the children with
same-sex parents had higher school
connectedness, significant at .05, and marginally
higher anxiety, which was not quite significant.
The t-test results shown in column B present
essentially the same results. School
connectedness, with a p-value of .015, is the only
comparison that is significant at .05, but anxiety
has a p-value of .07, that is, approaching but not
quite attaining significance at the
conventional .05 level.
Column C reports the observed mean value in
the Add Health full sample for WRP’s sample of
same-sex parents. The values in this column are
not exactly the same as those in column B. The
column B values were interpolated, which may
have introduced unknown error, but the most
likely source of the differences between the
columns is differences in missing data. The
present study computed mean values from all
non-missing cases for each outcome variable (for
most outcomes either 43 or 44 cases), but WRP
analyzed the variables in three structural groups;
if data were missing for any outcome variable in
the group, it was counted as missing for all
variables in the group. For most of the outcome
variables shown, this analytical decision
substantially reduced the number of cases on
which their mean value computations were
actually made. For depressive symptoms, for
example, W RP’s mean value of 10.93, shown in
column B, was based on 27 cases, while the
corresponding value shown in column C,
computed for the present study, is based on 43
cases. The values in column C, therefore, are
generally more accurate than those in column B,
although the differences are generally slight. For
only three variables are the p-values testing
mean difference higher in column C than in
column B. In the bottom four rows of Table 1,
WRP’s reported values are based on the highest
number of cases in their same-sex parents
sample (43 of 44), so the column B values are
most similar to column C for those outcomes; for
neighborhood integration the values are identical.
In column C no adolescent differences are
significant at .05, although school connectedness
is still significant at the .10 level. Likewise, no
difference is significant at .05 on any outcome for
any of the remaining columns of the table
(columns D, E, and F). For column F, showing
results for the 6 gay male parent couples, school
connectedness is also significant at .08,
suggesting that the results for this group could be
described as “very nearly identical” to those of
column B, but this does not seem to be the case
for column E, which shows the 17 actual same-
sex parent cases. For this group, school
connectedness is not significantly different from
the matched sample shown in column A, as is
the case for WRP’s findings for the full group of
44 alleged same-sex parent cases shown in
column B. Moreover, child GPA (grade point
average) is significant at .06, very close to
the .05 level, which is decidedly not the case for
column B. Perhaps WRP’s matched comparison
group for this sample of 17 ideal same-sex
parent cases was different than that for the full
sample of 44 cases.
Columns D and E, as noted, disaggregate the 44
cases shown in columns B and C into the 27
cases of misidentified opposite-sex parents and
the 17 clear cases of lesbian parents respectively.
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
8
Notably, as judged by p-value, column E has
more items that are closer to being significantly
different from column B than does column D (5
compared to 3), despite the fact that it has fewer
cases. GPA, depressive symptoms and anxiety
are much closer to significance in column E than
in column D. For the 17 ideal cases in column E,
all three variables measuring adolescent
psychological well-being (depressive symptoms,
self-esteem and anxiety) and family and
relationship processes (parental warmth, care
from adults and peers, autonomy and integration)
show less favorable results, but all three school
outcome variables (GPA, trouble in school and
school connectedness) show more favorable
results.
4.2.2 Replication with the corrected sample
Table 2 presents new means tests results for the
outcome variables in WRP 2004 after correcting
the same-sex parent sample to remove the 27
erroneous cases of opposite-sex parent partners
and applying the appropriate sample weights.
The corrected same-sex parents sample,
reported in column E, also includes 3 of the 6
cases of male same-sex parents, who were
verified by the same stricter screening
procedures used to verify the clear cases of
female same-sex parents, for a total sample of
20 clear cases of parenting same-sex partners.
The analyses presented in Table 2 generally
confirm the accuracy of WRP’s analysis
regarding significant differences by family type,
given their use of a small extract of unweighted
cases and a corrupted sample. At the same time,
the new findings shown demonstrate the
increased power of the corrected sample, when
analyzed using sample weights and survey
design features, to identify differences, both
advantageous and disadvantageous, for children
with same-sex parents.
As in Table 1, combined variables or multivariate
tests are ignored. In the absence of W RP’s
matched sample of opposite-sex parents,
Column A in the table reports the unweighted
mean value for each outcome variable from the
Add Health Core Sample. Columns B-E report
outcome values under various conditions, with
corresponding t-test results. To facilitate
comparison, column B repeats the replicated
findings from WRP 2004 already shown in Table
1, column B. Columns C and D report
respectively the replicated values and
significance test results from the unweighted and
weighted Add Health Full Sample. Column E
shows the results for the corrected category of
same-sex parents. Columns D and E, but no
other, adjust variance estimates for survey
design and weights, and thus present results that
may be inferred to the population in question.
Table 2 confirms several points made in the
critique above. For every variable in the table,
the standard errors reported by WRP, shown in
Column B, are larger, in most cases much larger,
than those of any other sample condition shown.
This confirms that, as discussed above, WRP
analyzed the matched groups of 44 parents each
independently, not as part of the Add Health
dataset. Columns C and D show mean values for
the WRP 2004 sample computed with
unweighted and weighted cases respectively.
Consistent with the warning provided in the
Guidelines for Analyzing Add Health Data [12],
for all but two variables the standard errors for
the unweighted values (Column C) are smaller
than the standard errors for the weighted values
(Column D). The mean values reported by WRP
2004 for the “lesbian parents” sample (Column
B), which is really composed primarily of
heterosexual parents, are, with two exceptions,
very similar to the mean value (unweighted) for
the Add Health Core Sample.
As already noted, WRP reported only one
significant difference by family type: children with
same-sex parents had significantly higher school
connectedness (than did the comparison group
of children with opposite-sex parents). Table 2
confirms this finding when comparing the
weighted cases of children with same-sex
parents to the mean of the full sample. In the
corrected sample (Column E), school
connectedness for children with same-sex
parents is even higher, with higher statistical
significance.
WRP did not find a significant difference for
grade point average by family type, but this is
also found to be significantly higher for the WRP
2004 sample when sample weights and
clustering are incorporated (Column D), and
even higher, with a more significant difference,
when the sample is corrected to include only
clear cases of same-sex parents (Column E). For
anxiety WRP reported results that were a third
larger for boys, and a sixth larger for girls, with
same-sex parents, with a large F-statistic (4.5)
for the difference by family type (4.5). However,
they reported that multivariate anova revealed no
significant effects, so they concluded that there
was no difference. Table 2 confirms this
conclusion for the full WRP 2004 sample of 44
cases.
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
9
In sum, when the original sample is corrected to
include only same-sex parents, the mean for
adolescents with those parents differs
significantly from their counterparts with
opposite-sex parents on three of the ten
outcomes examined: anxiety, grade point
average (GPA), and school connectedness. In
the next section, the inclusion of control variables
confirms and extends this finding.
4.2.3 Replicating control variables
Table 3 presents the results of multiple
regression models that include controls for child
sex, age, race, and adoption status, and parent
age, education and income, thus more closely
replicating WRP’s comparison of two groups
which were matched on the same characteristics.
Coefficients for control variables were significant
for all outcomes. When using the WRP 2004
sample of same-sex parents, the regression
models with controls found, just as WRP did, that
the only variable that is significantly different by
family type is school connectedness. In the
corrected sample, school connectedness, grade
point average, and anxiety all remain significantly
higher, as they were in Table 2, in the presence
of controls. In addition, after including controls
child autonomy is significantly lower for children
with same-sex parents. These findings confirm
and extend the findings of Table 2.
4.3 Step Three: Re-analysis
Forty percent of the same-sex partners reported
their marital status as married rather than as
unmarried partners. This is consistent with other
representative data such as the National Health
Interview Survey and the 2000 Census, where
many same-sex couples also indicated that their
partnership was a marriage prior to same-sex
marriage attaining legal status in any part of the
United States in 2004. While not legally
recognized marriages, these cases clearly reflect
a marital self-understanding, and the partners
may well have been married in a religious or
private ceremony during this era. Prior studies
have found that such couples may be plausibly
interpreted as reflecting many of the attributes of
marriage [2,15–17], thereby offering, as Reczek
and colleagues conclude, “our closest possible
representation of the current population of the
same-sex married” [17]. On Add Health the
married same-sex parents strongly reflect,
moreover, the most commonly-referenced
potential advantage of marriage for same-sex
parents: greater family stability. As discussed
below, the time children had resided with their
current set of parents averaged 4 years (SE 2.3)
with unmarried same-sex partners, but with
married same-sex partners, 10.4 years (SE 3.1).
Table 4, accordingly, reports the findings of a re-
analysis of the Add Health data, with the
corrected same-sex parent category expressed
in the Full Sample, by family type and marriage;
figures 1-6 illustrate selected effects. The table
presents the findings of logistic regression
models that impose the seven demographic
controls used by WRP. The reference category
for statistical tests is opposite-sex married
parents.
In Table 4, due to the sparseness of the data, the
57-point CES-D scale and related subscales are
expressed as dichotomous predictors divided at
the median of the distribution. It is important to
bear in mind that the resulting categories do not
predict for a psychological disorder or an
abnormal level of depressive symptoms.
Depressive symptoms are lower than average
(47.2% SE .89 are above average) for children
with opposite-sex married parents. Child
depressive symptoms are 9 points higher with
unmarried opposite-sex parents (56.0% SE 1.1)
and a striking 40 points higher with married
same-sex parents (87.7% SE 11). Among
children with unmarried parents, depressive
symptoms (50.4% SE 25) are lower with same-
sex parents than with opposite-sex parents,
though the difference is not statistically
significant. See Fig. 1. The same pattern can be
observed, only more strongly, on the CES-D
subscale for lack of positive affect (unhappiness).
Children with unmarried same-sex parents are
much less unhappy (34.0% SE 20) than children
with unmarried opposite-sex parents (56.9% SE
1.0), but children with married same-sex parents
are much more unhappy (94.9% SE 6) than are
children with married opposite-sex parents
(51.3% SE.86). See Fig. 2.
Negative interpersonal symptoms are lower
overall for children with same-sex parents,
suggesting that they are not subject to
widespread social rejection, or at least not as
much as are children with opposite-sex parents.
Nonetheless, children whose same-sex parents
are married are over twice as likely to have
above-average negative interpersonal symptoms
(22.7% SE 9) than are those whose same-sex
parents are unmarried (11.5% SE 8). See Fig. 3.
On the other hand, anxiety is significantly higher
for children with both unmarried and married
same-sex parents, although the latter are higher.
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
10
With marriage, child anxiety drops (from 4.65
SE .09 to 4.51 SE .05) with opposite-sex parents,
but rises (from 6.31 SE .77 to 7.1 SE 1.5) with
same-sex parents. See Fig. 4.
The proportion of children reporting daily
fearfulness or crying, compared to children with
married opposite-sex married parents (3.1%
SE .25), is moderately higher for children with
unmarried opposite-sex parents (4.4% SE .46)
and unmarried same-sex parents (5.4% SE 5.7),
but much higher—over ten times as high—for
children with married same-sex parents (32.4%
SE 25.2). Almost a third of children with same-
sex married parents reported feeling fearful or
crying daily. This difference is not significant in
Table 4, but (as discussed below) is highly
significant in the maximum likelihood models
after fitting control variables.
Fig. 1. Depressive symptoms by family type
and marriage
Fig. 2. Lack of positive affect by family type
and marriage
Fig. 3. Negative interpersonal relations by
family type and marriage
Fig. 4. Child anxiety by family type and
marriage
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
11
Unlike psychological well-being, both grades and
school connectedness are higher with same-sex
parents than with opposite-sex parents. Parental
warmth estimates are also slightly higher with
same-sex parents, though the difference is not
significant. Like the scales for negative
interpersonal relations and lack of positive affect,
perceived care from adults and peers is higher
for children with unmarried same-sex parents,
but lower for children with married same-sex
parents, than it is for the corresponding
categories of children with opposite-sex parents.
In all of these contrasts, however, the pattern of
higher well-being with unmarried same-sex
parents rather than married same-sex parents
continues to be observed.
Grade point average (GPA), for example, is
higher overall for children with same-sex parents,
but while GPA is lower with unmarried opposite-
sex parents (2.6 SE .02) than with married
opposite-sex parents (2.9 SE .02), it is higher
with unmarried same-sex parents (3.6 SE .31)
than with married same-sex parents (3.4 SE .12).
See Fig. 5.
Two variables in Table 4 measure family stability.
The length of time the adolescents have been
with in their current family relates to whether the
outcomes observed are due to the current
parents or may be the effect of residence with
former parents. Recall that average age is 15
years for the Add Health adolescent respondents.
Adolescents with opposite-sex married parents
have the longest duration with those parents, at
13 years. Average duration drops by about 2.5
years with unmarried opposite-sex parents (10.4
years SE .18) and married same-sex parents
(10.4 years SE 3.1), then plummets to only 4
years (SE 2.3) with unmarried same-sex parents.
By this measure, married same-sex parents are
much more stable, though child well-being is
generally lower, than are unmarried same-sex
parents.
The percentage of children who have undergone
one or more relational transitions from one set of
parents to another one, a related measure, is
lowest for children with opposite-sex married
parents and highest for those with same-sex
married parents; the latter is over four times the
size of the former. Almost all (83%-88% SE 11-
16) children with same-sex parents have
experienced at least one relational transition,
compared to under half (45% SE 1.3) of children
with unmarried opposite-sex parents and less
than a fifth (19% SE .75) of children with
opposite-sex married parents. By this measure,
married same-sex parents are a little less stable
than unmarried same-sex parents, though both
are much less stable than opposite-sex parents.
Fig. 5 Grade Point Average by Family Type
and Marriage
Fig. 6. Forced Sex by Family Type and
Marriage
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
12
Table 1. Replication of WRP’s analysis with alternative samples of same-sex parents: Add Health wave 1
A
B
C
D
E
F
44 opposite-
sex cases
(reported)
44 same-sex
cases (reported)
44 same-sex
cases (observed,
unweighted)
27 “real world”
cases
(unweighted)
17 “ideal”
cases
(unweighted)
6 same-sex
male parent
couples
(unweighted)
Mean
(SD)
Mean
(SD)
p > t
(A=B)
Mean
(SD)
p > t
(A=C)
Mean
(SD)
p > t
(A=D)
Mean
(SD)
p > t
(A=E)
Mean
(SD)
p > t
(A=F)
Depressive symptoms (CES-D) 9.67
(6.24)
10.93
(8.46)
.50 11.53
(8.10)
.25 10.70
(8.81)
.60 12.94
(6.79)
.11 13.33
(6.15)
.22
Self-esteem 4.04
(.62)
3.99
(.50)
.73 4.19
(.64)
.29 4.30
*
(.55)
.08 4.0
(.73)
.85 3.97
(.31)
.68
Anxiety (6 items only) .76
(.44)
.99
*
(.53)
.07 .85
(.62)
.45 .76
(.60)
.99 1.0
(.64)
.17 .56
(.51)
.40
GPA (grade point average) 2.80
(.78)
2.83
(.90)
.88 3.00
(.82)
.32 2.86
(.87)
.80 3.3
*
(.67)
.06 2.65
(.98)
.73
Trouble in school .95
(.73)
1.03
(.70)
.64 1.10
(.80)
.39 1.18
(.73)
.22 .97
(.92)
.94 .79
(.84)
.68
School connectedness 3.43
(.83)
3.92
**
(.81)
.015 3.73
*
(.71)
.096 3.75
(.76)
.12 3.70
(.62)
.20 3.72
*
(.20)
.08
Parental warmth 4.39
(.34)
4.27
(.54)
.22 4.23
(.59)
.13 4.30
(.49)
.41 4.11
(.73)
.15 4.4
(.35)
.99
Care from adults and peers 4.09
(.62)
4.04
(.69)
.72 4.05
(.69)
.27 4.12
(.62)
.84 3.94
(.80)
.50 4.17
(.62)
.77
Autonomy 5.44
(1.30)
5.19
(1.59)
.43 5.11
(1.47)
.84 5.30
(1.03)
.62 4.82
(1.96)
.24 5.67
(1.21)
.68
Neighborhood Integration 2.37
(.93)
2.21
(.91)
.42 2.21
(.91)
.42 2.26
(.86)
.62 2.13
(1.02)
.42 1.83
(.75)
.15
Columns A & B report interpolated results from WRP 2004 Table 2 (p. 1892), which are slightly different than those reported in the text. Except for column A and B all
statistics, including t-test comparisons, are based on the Add Health Wave 1 full sample (n=20,745):
*
t, P < 0.10;
**
t, P < 0.05;
***
t, P < 0.01;
****
t, P < 0.001. 4.54 Anxiety scale
is transformed to a 0-4 range
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
13
Table 2. Adolescent characteristics as a function of family type: Add Health wave 1
A
B
C
D
E
Add Health
core sample
(12,105)
WRP 2004 (reported) (44)
WRP 2004 observed
(unweighted)
WRP 2004 observed
(weighted)
Corrected SS parents
sample (weighted)
Mean
(SE)
Mean
(SD)
p > t
(ss=os)
Mean
(SE)
p > t
(ss=os)
Mean
(SE)
p > t
(ss=os)
Mean
(SE)
p > t
(ss=os)
Depressive symptoms
(CES-D)
10.91
(.137)
10.93
(8.46)
.50 11.53
(1.24)
.91 10.43
(.940)
.54 11.06
(1.48)
.96
Self-esteem 4.12
(.01)
3.99
(.50)
.73 4.19
(.10)
.40 4.26
(.14)
.28 4.10
(.23)
.94
Anxiety (6 items only) .76
(.01)
.99
*
(.53)
.07 .85
(.62)
.28 .92
(.11)
.16 1.12
***
(.14)
.01
GPA 2.83
(.02)
2.83
(.90)
.88 3.00
(.15)
.14 3.16
*
(.19)
.08 3.49
****
(.21)
.002
Trouble in school 1.06
(.01)
1.03
(.70)
.64 1.10
(.12)
.63 1.02
(.14)
.78 .77
(.24)
.24
School connectedness 3.61
(.01)
3.92
**
(.81)
.015 3.73
(.11)
.20 3.92
**
(.13)
.02 4.04
***
(.16)
.009
Parental warmth 4.30
(.01)
4.27
(.54)
.22 4.23
(.09)
.63 4.44
(.09)
.13 4.50
(.17)
.23
Care from adults and peers 4.06
(.01)
4.04
(.69)
.72 4.05
(.11)
.91 4.17
(.17)
.71 4.25
(.26)
.44
Autonomy 5.11
(.05)
5.19
(1.59)
.43 5.11
(.22)
.84 4.71
(.35)
.23 4.16
(.64)
.13
Neighborhood Integration 2.24
(02)
2.21
(.91)
.42 2.21
(.14)
.98 2.12
(.20)
.54 1.89
(.43)
.41
Column B reports interpolated results from WRP 2004 Table 2 (p. 1892), which are slightly different than those reported in the text. To facilitate comparison standard deviations
are converted to standard errors. Statistics for columns C, D, and E, including t-test comparisons, are based on the Add Health Wave 1 full sample (n=20,745):
*
t, P < 0.10;
**
t,
P < 0.05;
***
t, P < 0.01;
****
t, P < 0.001
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
14
Table 3. Multiple regression coefficients predicting child characteristics by family type: Add
Health Wave 1
SS parents
WRP 2004 observed
(weighted)
Corrected SS parents
sample
Coeff
P>t
Coeff
P>t
Depressive symptoms (CES-D)
-.428
.31
.058
.96
Self-esteem .059 .41 .043 .85
Anxiety (6 items only) .259 .48 1.70
*
.08
GPA .089 .37 .430
***
.004
Trouble in school -.043 .51 -.232 .30
School connectedness .117
*
.06 .407
***
.007
Parental warmth .070 .16 .222 .16
Care from adults and peers .007 .93 .134 .58
Autonomy -.27 .13 -1.27
**
.03
Neighborhood Integration -.081 .42 -.325 .43
Shown are OLS regression models controlling for child sex, age, race (white/nonwhite), and adoption status;
parent age and education (college degree or not), and family income.
*
t, P < 0.10;
**
t, P < 0.05;
***
t, P < 0.01;
****
t,
P < 0.001
The remaining variables in Table 4 explore
different issues of sexual development and family
formation. Six percent of adolescents with
opposite-sex married parents reported that they
have ever been romantically or sexually attracted
to someone of the same sex. This proportion
rises to 8 percent with unmarried opposite-sex
parents, then to much larger estimated
proportions with same-sex parents, although the
differences are not statistically significant.
Despite apparently higher rates of same-sex
attraction, no child with same-sex parents
reported ever having had a same-sex romantic
partner. Adolescents with same-sex parents
were also about half as likely to have ever had
sexual intercourse. In an item taken from the
Wave III follow-up, those with unmarried same-
sex parents were less likely, and those with
married same-sex parents more likely, to be
divorced or cohabiting with an unmarried partner
six years after the initial Add Health interview.
Over half of the children with married same-sex
parents were divorced or cohabiting after six
years.
The last two lines of Table 4 report findings on
the sensitive topic of child sex abuse. To
increase accuracy, adolescents entered their
answers to these sensitive questions
anonymously into a laptop computer in response
to recorded questions they heard using
earphones. Adolescents who had ever had
sexual intercourse were given a series of follow-
up questions that included being asked about
forced sex. Males were asked if they had ever
physically forced someone to have sexual
intercourse; females were asked if they had ever
been physically forced to have sexual intercourse.
This is the only item examined in the present
study where the question varied by gender. Of
adolescents who had ever had sexual
intercourse, 10% to 12% (SE .73-.92) of those
with opposite-sex parents reported having been
forced (or forcing someone) to have sexual
intercourse. This proportion doubles with same-
sex unmarried parents (24% SE 23), and almost
triples again with same-sex married parents.
Over two-thirds (71% SE 30) of the children with
same-sex married parents who had ever had
sexual intercourse reported that they had been
forced to have sex against their will at some point.
All the “yes” responses for this group are from
female adolescents, meaning that these are all
reports of being forced, not forcing someone else,
to have sex relations. In fact, strikingly, every
sexually active female adolescent living with
married same-sex parents (which are all lesbian
parent couples) responded “yes” to having
experienced forced sex. On the other hand, as
already noted this group of adolescent females
were only about half as likely to have ever had
sexual intercourse (15%) than were those with
married opposite-sex parents (32%), though the
difference is not statistically significant; and this
question does not preclude the possibility that
they had experienced date rape or peer sexual
abuse.
The final item in Table 4, however, clarifies that
much of the sex abuse reported did occur in the
family and confirms that the prevalence of abuse
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
15
was much higher with married same-sex parents
than in the other family types. This question, a
retrospective item from a subsequent wave of
Add Health, was asked of all respondents, not
just those who had ever had sexual intercourse.
The question asks whether the responding
adolescent had ever, prior to the sixth grade,
been forced to give or receive a sexual touch or
to have intercourse by a parent or caregiver. A
total of 38% (SE 14) of respondents with married
same-sex parents reported that they had
experienced such abuse, compared to much
smaller proportions (0-7% SE 0-.6) of the other
three categories of marriage and family type.
Table 5 sharpens the contrasts by imposing
control variables to assess whether the
differences between the groups can be the result
of demographic differences rather than marriage
or family type. The table reports linear regression
predictors adjusted for child age, sex and race,
and parent education and income, i.e., the same
variables on which WRP matched their samples.
Most of the contrasts show little or no change,
and few are significantly reduced, after
accounting for these control conditions. For
same-sex married parents, the following
contrasts are stronger or have higher statistical
significance in the regression models with
controls: anxiety, parental warmth, child’s time in
current family, forced sex and parent sex abuse.
The following are lower or have lower
significance: depressive symptoms, interpersonal,
lack of positive affect, and care from adults and
peers. None of the differences by family type for
married persons is rendered insignificant after
adjusting for controls.
As additional scrutiny to support or withhold
further confidence in these findings, the mean
and regression contrasts reported in Tables 3
and 5 were also estimated by maximum-
likelihood procedures to assess the possibility of
small-sample bias. Table 6 shows the results for
the smallest category, married same-sex parents.
The reference category for all contrasts is
married opposite-sex parents. The first two
columns re-present for convenience the mean
and regression results already reported in Tables
2 and 3.
The remaining two columns predict the same
contrasts using two forms of logistic regression.
The third column shows the result of canonical
binary logistic regression employing case
weights and survey design clusters. The results
generally, though not always, confirm the
consistent results of the linear analyses shown in
the first two columns. Since logistic regression
may be biased when one of the comparison
groups are very sparse, column four reports the
results of a bias-adjusted logistic regression
designed for rare events estimation. Developed
by mathematician David Firth, this form of logistic
regression penalizes the log-likelihood so as to
produce unbiased estimates even when one
category is very sparse [18]. However, the Firth
method cannot make use of the sample weights
and clustering used on Add Health. Thus, while
the resulting point estimates for the Firth logistic
regression are probably less accurate than those
of regular logistic regression, when the
significance probability is very different between
the two methods, we may suspect that the
canonical estimates are biased, thus providing
greater confidence that they are not biased in the
alternative condition. Taking .25 or greater as
“very different”, and confining ourselves to cases
where the decision on the null hypothesis would
be changed by the difference, in Table 6 this is
the case for “Depressive symptoms”, “GPA”, and
“Divorced/cohabiting at age 19-25”. While all of
these contrasts are significant, and the first two
highly significant, in the linear analyses, this
comparison suggests that these findings may not
be as robust as other findings in the table. On the
other hand, both logistic estimates are highly
significant for the contrast for “Daily fearfulness/
crying”, which is substantively large but not
significant in the linear models.
In general, contrasts that are confirmed using
more of the methods shown in Table 5 are likely
more robust and merit higher confidence. By this
test, the strongest finding shown is for parental
sex abuse, which is large and significant by all
four methods. All of the psychometric contrasts
are consistent over three methods, as is GPA,
school connectedness, later divorce/cohabitation,
and forced sex. While no finding in the table is
invalidated by these additional comparisons,
those with more consistent findings may merit
additional confidence.
5. DISCUSSION
Almost all scholarly and policy consideration of
same-sex marriage has assumed that marriage
between partners of the same sex would result in
improved outcomes for children, just as marriage
generally does for children with opposite-sex
parents. This presumption is so widespread and
so strong that the prospect of improved child
well-being has been cited as one of the primary
justifications for regularizing same-sex marriage.
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
16
Table 4. Adolescent characteristics as a function of family type and marriage, showing
unadjusted mean values: Add Health Waves 1 and 3
Opposite-sex parents
Same-sex parents
Unmarried
Married
parents
Unmarried
Married
Mean
(SE)
p>t
(OS
Marr)
Mean
(SE)
Mean
(SE)
p>t
(OS
Marr)
Mean
(SE)
p>t
(OS
Marr)
Psychological well-being
Depressive symptoms (CES-D)
- percent above average
56.0
****
(1.1)
.00 47.2
(.89)
50.4
(24.6)
.90 87.7
****
(11.1)
.00
2CES-D Interpersonal –
People unfriendly or disliked
you
- percent above average
50.0
****
(1.0)
.00 44.8
(.71)
11.5
(8.4)
.19 22.7
****
(9.0)
.00
CES-D Lack of Positive Affect –
Not hopeful, happy, joyful
- percent above average
56.9
****
(1.0)
.00 51.3
(.86)
34.0
(19.7)
.38 94.9
****
(6.2)
.00
Anxiety 4.65
*
(.09)
.09 4.51
(.05)
6.31
**
(.77)
.02 7.10
*
(1.45)
.08
Daily fearfulness/crying (%) 4.4%
***
(.46)
.004 3.1%
(.25)
5.4%
(5.7)
.69 32.4%
(25.2)
.25
School outcomes
GPA 2.64
****
(.02)
.00 2.91
(.02)
3.59
**
(.31)
.04 3.37
****
(.12)
.00
School connectedness 3.51
****
(.02)
.00 3.66
(.01)
4.10
(.28)
.13 3.98
****
(.03)
.00
Family process
Parental warmth 4.21
****
(.02)
.00 4.34
(.01)
4.59
(.24)
.29 4.41
(.22)
.75
Care from adults and peers 3.99
****
(.02)
.00 4.09
(.01)
4.64
***
(.18)
.003 3.78
****
(.08)
.00
Family stability
Child’s time in current family
(years)
10.35
****
(.18)
.00 13.03
(.12)
4.01
****
(2.3)
.00 10.36
(3.1)
.40
Percent child transitions 45.0%
****
(1.3)
.00 18.5%
(.75)
83.0
****
(16.1)
.00 88.0
****
(10.9)
.00
Sexual development/identity
Same-sex attraction 7.5%
***
(.53)
.001 5.5%
(.39)
23.2%
(17.5)
.31 19.0%
(9.6)
.16
Ever same-sex romantic partner 1.4%
****
(.20)
.000 .9%
(.13)
0%
****
(0)
.00 0%
****
(0)
.00
Ever sexual intercourse? 46.3%
****
(.02)
.00 32.7%
(.02)
27.8%
(.19)
.31 15.7%
(.15)
.22
Divorced/Cohabiting/ed at age
19-25
47.9%
****
(.02)
.00 36.2%
(.01)
35.2%
(.27)
.97 57.7%
**
(.11)
.047
(If ever intercourse): Ever
physically forced to have sex
against your will? - % yes
12.2%
****
(.92)
.00 10.0%
(.73)
23.5%
(23.1)
.31 70.5%
**
(29.7)
.04
Experienced sex abuse by
parent
6.8%
(.60)
.00 3.5%
(.33)
0%
****
(0.0)
.00 37.8%
**
(14.3)
.02
Unmarried includes single never married. Reference category for t tests is opposite-sex married parents. T-test results:
equality of means
*
t, P < 0.10;
**
t, P < 0.05;
***
t, P < 0.01;
****
t, P < 0.001. CES-D scales presented are not predictive of
psychological disorder
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
17
Table 5. Adolescent characteristics as a function of family type and marriage, showing
adjusted regression predictors: Add Health Waves 1 and 3
Opposite-sex parents
Same-sex parents
Unmarried
Married
(Ref)
Unmarried
Married
Coeff.
(95% CI)
P>t Coeff.
(95%
CI)
P>t Coeff.
(95% CI)
P>t
Psychological well-being
Depressive symptoms
(CES-D)
- above vs. below average
.056
****
(.03-.08)
.000 -- .030
(-.4-.4)
.89 .361
***
(.10-.62)
.006
CES-D Interpersonal –
People unfriendly or disliked
you
- percent above average
.043
****
(.02-.07)
.000 -- -
.324
****
(-.48-
.17)
.19 -.253
**
(-.47--.03)
.024
CES-D Lack of Positive Affect
Not hopeful, happy, joyful
- percent above average
.031
**
(.004-.06)
.025 -- -.173
(-.51-
.17)
.31 .473
****
(.31-.63)
.000
Anxiety .019
(-.01-.05)
.16 -- .279
****
(.16-
.40)
.000 .367
****
(.27-.46)
.000
Daily fearfulness/crying (%) .007
(-.003-.02)
.16 -- .010
(-.10-
.12)
.87 .303
(-.23-.83)
.26
School outcomes
GPA -.078
****
(-.11--.04)
.000 -- .287
****
(.25-
.33)
.000 .208
****
(.12-.30)
.000
School connectedness -.059
****
(-.09--.03)
.000 -- .338
****
(.23-
.45)
.000 .391
****
(.35-.43)
.000
Family process and
stability
Parental warmth -.036
***
(-.06--.01)
.005 -- .082
(-.27-
.44)
.65 .357
****
(.16-.56)
.001
Care from adults and peers -.055
****
(-.08--.03)
.000 -- .357
****
(.23-
.48)
.000 .002
(-.39-.39)
.99
Child’s time in current
family
(years)
-2.53
****
(-2.9--2.2)
.000 -- -
8.01
****
(-12.6-
-3.4)
.001 -5.01
*
(-10.6-0.6)
.08
Percent child transitions .246
****
(.22-.27)
.000 -- .655
****
(.28-
1.0)
.001 .729
****
(.47-.99)
.000
Sexual
development/identity
Same-sex attraction .022
****
(.01-.04)
.001 -- .195
(-.15-
.53)
.26 .138
(-.06-.34)
.18
Ever same-sex romantic
partner
.004
(.00-.01)
.14 -- -
.011
****
(-.02--
.01)
.000 -.012
****
(-.02--.01)
.000
Ever sexual intercourse? .102
****
(.07-.13)
.000 -- .096
(-.12-
.38 -.222
(-.56-.11)
.19
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
18
Opposite-sex parents
Same-sex parents
Unmarried
Married
(Ref)
Unmarried
Married
.31)
Divorced/Cohabiting/ed at
age 19-25
.094
****
(.06-.13)
.000 -- .042
(-.29-
.37)
.80 .247
**
(.05-.45)
.016
(If ever intercourse): Ever
physically forced to have sex
against your will? - % yes
.013
(-.01-.04)
.26 -- .068
(-.39-
.53)
.77 .576
**
(.09-1.0)
.021
Experienced sex abuse by
parent
.031
****
(.02-.05)
.000 -- -
.033
****
(-.05--
.02)
.000 .387
***
(.11-.66)
.007
Unmarried includes single never married. Reference category for coefficients is opposite-sex married (column two). T-
test results: significance of coefficient * t, P < 0.10; **t, P < 0.05; ***t, P < 0.01; ****t, P < 0.001. Predictors are adjusted
for age, sex, race, adoption status, family income, and parent age and education. CES-D scales presented are not
predictive of psychological disorder.
Table 6. Outcomes for same-sex married under various model assumptions:
Add health wave 1
Method Unadjusted
Mean/Percent
(no controls)
OLS
Regression
(with
controls)
Logistic
regression
(with controls)
Firth bias-
adjusted
logistic
regression
(with controls)
Mean or
Percent
P>t
OR P>t OR P>t OR P>t
Depressive symptoms
(CES-D)
87.7%
****
.00 .36
***
.006 6.36
*
.10 1.90 .41
CES-D Interpersonal 22.7%
****
.000 -.25
**
.024 .29
**
.067 .27 .15
CES-D lack of positive affect 94.9%
****
.000 .47
****
.000 19.3
**
.031 3.4 .19
Anxiety 7.10
*
.08 .37
****
.000 19.1
**
.011 3.6 .17
Daily fearfulness/crying (%) 32.4% .25 .30 .26 15.6
**
.043 12.1
***
.002
GPA 3.37
****
.000 .21
****
.000 7.4
*
.064 2.2 .40
School connectedness 3.37
****
.000 .39
****
.000 -- -- 12.0
*
.089
Parental warmth 4.41 .75 .36
***
.001 8.6
*
.086 3.4 .18
Care from adults and peers 3.78
****
.00 .002 .99 1.07 .94 .89 .87
Same-sex attraction 19.0% .16 .138 .18 3.96
*
.058 3.6 .16
Ever sexual intercourse? 15.7% .22 -.22 .19 .30 .37 .83 .83
Divorced/Cohabiting/ed at
age 19-25
57.7%
**
.047 .25
**
.016 3.02
***
.009 1.8 .47
(If ever intercourse): Ever
physically forced to have sex
against your will? - % yes
70.5%
**
.04 .57
**
.021 23.9
***
.002 10.3 .106
Experienced sex abuse by
parent
37.8%
**
.02 .39
***
.007 13.9
***
.007 7.7
**
.034
All models shown included controls for child sex, age, race (white/nonwhite), and adoption status; parent age and
education (college degree or not); and family income. Reference category for tests is opposite-sex married, except for
bias-adjusted models, which contrast same-sex married with all other. For dichotomous models outcome variables
were transformed to dichotomies at the distribution median.
*
t, P = < 0.10;
**
t, P < 0.05;
***
t, P < 0.01;
****
t, P < 0.001
The evidence presented in Table 4 calls that
presumption sharply into question. On every
measure, well-being for children with same-sex
parents is lower if those parents are married than
if they are not. Figs. 1-6 illustrate the effect,
showing findings from Table 4. Residing with
married rather than unmarried parents of the
same sex is associated with substantially
increased depressive symptoms, anxiety and
daily distress, and lower educational
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
19
achievement and school connectedness. The
extremely high lack of positive affect—lack of
hopefulness, happiness, a positive affirmation of
life—among children with married same-sex
parents, but low lack of positive affect among
children with unmarried same-sex parents, is
particularly notable.
To be sure, not all outcomes for children with
same-sex parents in these data are negative. In
the corrected sample reported in Table 3, four
significant differences are visible for children with
same-sex parents. Two of the differences related
to school performance—higher grade point
average and school connectedness—are
advantageous, consistent with Rosenfeld’s
(2010) finding that children with same-sex
parents progress normally through school. The
other two differences report lower outcomes on
two psychosocial measures—anxiety and
autonomy—consistent with studies that have
found that children with same-sex parents suffer
higher emotional distress [9,10]. The positive
“differences”, however, follow the same pattern
as do the negative psychological “differences”
with respect to marriage, i.e., they are more
positive for children with unmarried, rather than
married, same-sex parents. For example, the
mean grade point average of 3.6 for those
children with same-sex parents who are
unmarried drops to 3.4 if the parents are married;
although both of these numbers are higher than
corresponding means for children with opposite-
sex parents. Parental warmth and perceived care
from adults and peers are mixed, higher among
children with unmarried same-sex parents, but
lower for children with married same-sex parents,
than they are for children with opposite-sex
married parents.
In the absence of further information,
interpretation of these mixed results is
necessarily speculative. One possible
explanation for the co-presence of negative
psychological effects with positive educational
outcomes is that same-sex attracted persons,
and hence their children, may be more intelligent
than the general population. A similar co-
existence, of higher average incomes despite
increased psychological distress, has been well
established for the population of same-sex
attracted adults. It is also possible that the
negative and positive effects are partitioned,
each manifesting in a different portion of the
population in question.
Another possible explanation is consistent with
the recognition that, for the children with same-
sex parents, the relatively positive outcomes, like
school progress, family warmth and even
interpersonal perceptions, are more public
matters known to peers and community while the
negative psychological effects and child abuse
tend to be private and hidden. Previous research
has noted the tendency for same-sex parents to
minimize negative features in accounts of their
children’s lives [19,20]. For example, Malmquist
and Nelson, analyzing 96 lesbian mothers’
counterfactual descriptions of experiences with
maternal and parenting healthcare professional
as “just great”, observed that political concerns
shaped their rhetorical accounts: “at stake was
the risk of feeding opponents of lesbian
parenthood with arguments they could use
against these families, namely that it would be
harmful for any child to be brought up in a two-
mother family. Instead, the unproblematic
journey, a ‘just great’ story, was stressed,
highlighted and emphasized over and over
again”. Thus “when our interviewees claimed
their ‘just great’ stories, despite their descriptions
of inadequate encounters, they were accounting
for their creditability as competent parents” (20).
Moreover, just as parents have been reluctant to
supply negative accounts, researchers have
been reluctant to demand or acknowledge them
[21]. Parental bias of this sort could be avoided
or reduced by a greater use of third-party reports,
such as those of teachers, or, as Allen
recommends [3], the avoidance of subjective
reports in favor of more standardized, objective
measures of child well-being.
Lopez and Edelman, in a volume of qualitative
reports from children raised by same-sex couples,
have critiqued the “no differences” research on
just these grounds. “[S]ocial-science research
that has ostensibly shown positive “outcomes” for
children raised by same-sex couples… are really
just measurements of what adults want from
children so the adults look good: Does the child
have good grades? Does the child look happy in
photographs. …? Is the child well-adjusted,
healthy, a good athlete, well liked by his peers,
…? In other words, : Do children in same-sex
couple’s homes turn out the way gay people
want them to, so that gay people look good to
straight people” [22]? In support of this point, it is
striking that few studies (to my knowledge, only
four) in the no differences” literature have
employed standard psychometric measures of
emotional distress such as the CES-D or the
Strengths and Difficulties Questionnaire [23], and
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
20
no study has asked about parental child abuse.
If politically aware concern for demonstrably
positive child outcomes is as pervasive as these
accounts suggest, it is conceivable that same-
sex parents could also disproportionately
emphasize such demonstrable achievement in
their children, leading to just the kind of mixed
results observed in the Add Health data.
Increased family stability is often cited as a likely
benefit of same-sex marriage, but these findings
also call into question the premise of that
argument. Stability leads to more positive child
outcomes with opposite-sex partners, but it
appears to have the opposite effect for children
with same-sex parents. As Table 4 shows,
children whose same-sex parents were married
had been with that particular set of parents over
2.5 times longer, at over ten years on average,
than had children with unmarried same-sex
parents, at about four years on average.
Marriage did bring greater stability, but stability
did not bring better child outcomes: married
same-sex parents were much more stable,
though child well-being was generally lower, than
were unmarried same-sex parents. Similarly, the
proportion of children who had undergone at
least one transition from one set of parents to
another, such as in a divorce and remarriage,
was at least four times higher, at 83% and 88%
for unmarried and married same-sex parents
respectively, than it was for opposite-sex married
parents, at 19%. Such transitions are
experienced by children as traumatic, generally
impeding their well-being and development.
Perhaps the substantially higher rate of
transitions with same-sex parents, estimated at
even somewhat higher if they are married, may
help to account for the relatively lower child well-
being with married same-sex parents.
Multivariate models suggest that the effects of
tenure, transitions and marital status are largely
independent, although further research is
necessary to clarify the relationship of these
factors.
In sum, from the evidence presented in this
paper, it does not appear that the operational
benefits of marriage that accrue to opposite-sex
couples are severable from the man-woman
relationship. It may be that the kind of functional
thinking that underlies the argument that the two
forms of marriage relationship are analogous is
mistaken, and the beneficial factors that are
observed in man-woman marriage--greater
stability, financial resources, relational security—
do not float free in a manner that can be
independently conveyed to another kind of
relationship.
6. LIMITATIONS
Despite the signal strengths of Add Health as a
large nationally representative dataset, and
notwithstanding the strong significance for
contrast effects reported above, due to the small
sample sizes involved, the findings of this study
should be considered only provisional and
exploratory until and unless they are confirmed
by further research. In particular, the findings
presented in Table 4 and related analyses are
based on very small or sparse categories and
should not be considered definitive without
corroboration. Although Add Health enables
longitudinal analysis, this study examined data
from only one wave, and thus, as with any cross-
sectional data, causal inference is not possible.
The findings presented in this study are focused
on an assessment of measures presented in
prior studies, and should not be taken as
presenting a comprehensive profile of parenting
outcomes.
7. CONCLUSION
Contrary to the expectations prompted by the “no
differences” literature and related ideologies,
harm for children with same-sex parents does
not appear to be attributable to prior
heterosexual relationships, lower stability,
relational commitment, or higher stigma among
same-sex parents. In the data observed in this
study, the greatest harm for children with same-
sex parents came from the most stable and most
marital family arrangements. This unexpected
harm was present despite warm and loving
parents who promoted positive school outcomes,
but also may be related to higher rates of abuse.
Recent first-person narrative accounts of growing
up with same-sex parents have presented a
complex image of harm despite positive parental
qualities that is very similar to the impression
suggested by these findings [22,24,25].
The present study has re-examined some of the
strongest evidence adduced in support of the no
differences thesis, concluding that, when re-
analyzed in a manner that could show
differences if they existed, such differences are
manifestly present. As noted in the introduction,
a steady drumbeat of dozens of studies based on
small, non-random samples has been celebrated
by the American social science establishment as
definitive proof that having same-sex parents is
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
21
innocuous for child well-being. In the face of
mounting evidence to the contrary, the American
Psychological Association continues to claim:
“Not a single study has found children of lesbian
or gay parents to be disadvantaged in any
significant respect relative to children of
heterosexual parents” [26]. The present study
definitively demonstrates that statement to be
false.
To those convinced that the no differences thesis
is true, the evidence presented in this study is
unexpected and possibly inconvenient. Whether
future evidence upholds, modifies or rebuts these
findings, they suggest that much of the received
social science wisdom about such relationships
is mistaken, and we have just begun to try to
understand the effect on children of having two
parents of the same sex.
COMPETING INTERESTS
Author has declared that no competing interests
exist.
REFERENCES
1. Manning WD, Fettro MN, Lamidi E. Child
well-being in same-sex parent families:
Review of Research Prepared for
American Sociological Association Amicus
Brief. Popul Res Policy Rev. 2014;33(4):
485.
2. Rosenfeld MJ. Nontraditional families and
childhood progress through school.
Demography. 2010;47(3):755-75.
3. Allen DW. High school graduation rates
among children of same-sex households.
Rev Econ Househ. 2013;11(4):635-58.
4. Wainright JL, Russell ST, Patterson CJ.
Psychosocial adjustment, school
outcomes, and romantic relationships of
adolescents with same-sex parents. Child
Dev. 2004;75(6):1886-98.
5. Wainright JL, Patterson CJ. Delinquency,
victimization, and substance use among
adolescents with female same-sex
parents. J Fam Psychol. 2006;20(3):
526–30.
6. Wainright JL, Patterson CJ. Peer relations
among adolescents with female same-sex
parents. Dev Psychol. 2008;44(1):117.
7. Patterson CJ. Lesbian and gay parents
and their children: A social science
perspective. Contemporary Perspectives
on Lesbian, Gay, and Bisexual Identities
[Internet]. Springer. 2009;141–82. [cited
2015 Mar 27].
Available:http://link.springer.com/chapter/1
0.1007/978-0-387-09556-1_6
8. Fedewa AL, Clark TP. Parent practices
and home-school partnerships: A
differential effect for children with same-
sex coupled parents? J GLBT Fam Stud.
2009;5(4):312-39.
9. Regnerus M. How different are the adult
children of parents who have same-sex
relationships? Findings from the New
Family Structures Study. Soc Sci Res.
2012;41(4):752-70.
10. Sullins DP. Emotional Problems among
Children with Same-sex Parents:
Difference by Definition. Br J Educ Soc
Behav Sci. 2015;7(2):99-120.
11. Allen DW, Pakaluk C, Price J.
Nontraditional families and childhood
progress through school: A Comment on
Rosenfeld. Demography. 2013;50(3):
955–61.
12. Chen P, Chantala K. Guidelines for
analyzing add health data [Internet].
Carolina Population Center, University of
North Carolina at Chapel Hill; 2014.
Available:http://www.cpc.unc.edu/projects/
addhealth/data/guides/wt-guidelines.pdf
13. Radloff LS. The CES-D scale: A self-report
depression scale for research in the
general population. Appl Psychol Meas.
1977;1(3):385-401.
14. National longitudinal survey of adolescent
health. In-home questionnaire code book
(Wave I). Carolina Population Center,
University of North Carolina at Chapel Hill;
1995.
15. Liu H, Reczek C, Brown D. Same-sex
cohabitors and health: The role of race-
ethnicity, Gender, and Socioeconomic
Status. J Health Soc Behav. 2013;
54(1):25–45.
16. Reczek C, Liu H, Brown D. Cigarette
Smoking in Same-Sex and Different-Sex
Unions: The Role of Socioeconomic and
Psychological Factors. Popul Res Policy
Rev. 2014;33(4):527-51.
17. Reczek C, Liu H, Spiker R. A Population-
Based Study of Alcohol Use in Same-Sex
and Different-Sex Unions. J Marriage Fam.
2014;76(3):557-72.
18. Firth D. Bias Reduction of Maximum
Likelihood Estimates. Biometrika. 1993;
80(1):27–38.
19. Clarke V, Kitzinger C, Potter J. Kids are
just cruel anyway: Lesbian and gay
parents’ talk about homophobic bullying. Br
J Soc Psychol. 2004;43:531-50.
Sullins; BJESBS, 11(2): 1-22, 2015; Article no.BJESBS.19337
22
20. Malmquist A, Nelson KZ. Efforts to
maintain a “just great” story: Lesbian
parents’ talk about encounters with
professionals in fertility clinics and
maternal and child healthcare services.
Fem Psychol. 2014;24(1):56–73.
21. Stacey J, Biblarz TJ. (How) does the
sexual orientation of parents matter? Am
Sociol Rev. 2001;66(2):159–83.
22. Lopez R, Edelman R, editors. Jephthah’s
Daughters. Northridge, CA: International
Children’s Rights Institute. 2015;486.
23. Goodman R. Psychometric Properties of
the Strengths and Difficulties
Questionnaire. J Am Acad Child Adolesc
Psychiatry. 2001;40(11):1337–45.
24. Edwards J. Like Mother, Like Daughter?:
The effects of growing up in a homosexual
home. Xulon Press. 2001;206.
25. Stefanowicz D. Out From Under: The
Impact of Homosexual Parenting.
Redemption Press. 2014;116.
26. Patterson CJ. Lesbian and gay parents
and their children: Summary of research
findings [Internet]. American Psychological
Association; 2005.
Available:http://www.apa.org/pi/lgbt/resour
ces/parenting-full.pdf
_________________________________________________________________________________
© 2015 Sullins; This is an Open Access article distributed under the terms of the Creative Commons Attribution License
(http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium,
provided the original work is properly cited.
Peer-review history:
The peer review history for this paper can be accessed here:
http://sciencedomain.org/review-history/10443
... While few of those results were significant statistically, they represented medium to large effect sizes. Sullins [62,63] in an analysis of ADD HEALTH data found that as of WAVE IV, adolescents with same-sex parents had higher rates of depression (d = .85), suicidal ideation (d = .97), ...
Article
Full-text available
Research has been ongoing regarding homosexuality and same-sex parenting for over fifty years, yet a recent (2020) review of the literature stated that the field was still in its infancy. Here research on same-sex parenting and child outcomes is reviewed, considering how our knowledge base has or has not changed in various topical areas. Consistently, the predominant paradigm has been the "no differences" hypothesis, meaning that there are no differences between same-sex and opposite-sex families, parents, or children. This review finds that population estimates of same-sex families have changed over time and enough empirical evidence has accumulated to challenge the "no differences" paradigm in some areas, while in other areas, the paradigm has held or the available evidence has been mixed.
... He concluded that full biological parentage, a modality not possible for same-sex parents, does influence child emotional problem outcomes. Sullins [17] also reevaluated three studies by Wainright & Patterson [18][19][20] on 44 SSC and found what he considered major inaccuracies in these studies. In particular, 27 out of the 44 adolescents were in fact living with opposite-sex parents and, after correction, the 17 adolescents truly living in a SSF fared significantly worse than did their counterparts. ...
... Since same-sex couples comprise less than one percent of all couples in a given population, including the population of "only 26.6 percent of same-sex female couples and 22.2 percent of same-sex male couples are correctly coded" (Black et al., 2007, p. 10). Sullins, who checked Wainright et al.'s sample of same-sex parent cases identified on Add Health using a secondary sex verification that Wainright had not consulted, found that 61% of the cases identified as "same-sex parents" in that study consisted of different-sex parent partners (Sullins, 2015). ...
... (In a previous study I examined tenure and family transitions in these data in more detail, with similar, and statistically significant, results.) [12] Thus, contrary to what Frank alleges, the lack of control for stability or tenure does not overstate, and may understate, negative child outcomes with same-sex parents. ...
Article
Full-text available
The table also addresses Frank’s concern that not accounting for shorter stability may overstate child disadvantages with same-sex parents. The time children have been in the care of both current parents is indeed much shorter with same-sex than opposite-sex parents, but the effect of this difference is not what Frank speculates. The table compares age-adjusted odds ratios by family type, by length of tenure with both current parents, for the three primary affective outcomes I examined. Both-parent tenure (ranging from 0 to 20 years) is grouped into four five-year categories; the reference category is five years or fewer. For children with opposite-sex parents, longer tenure with the current parents is associated with lower odds of depression, anxiety, and thoughts of suicide; but the evidence does not support a similar conclusion for same-sex parents, as Frank assumes. For children with same-sex parents, odds of suicide ideation are unchanged by both-parent tenure and odds of depression and anxiety are, if anything, higher with longer tenure. (In a previous study I examined tenure and family transitions in these data in more detail, with similar, and statistically significant, results.) [12] Thus, contrary to what Frank alleges, the lack of control for stability or tenure does not overstate, and may understate, negative child outcomes with same-sex parents.
Article
Full-text available
While a few have argued that social science has been subject to progressive biases, others have discounted such ideas. However, no one has yet performed empirical tests over a large range of studies for such possible bias, which we label macro-level social desirability (MLSD). Combining the results from fifty-nine empirical studies that assessed rates of nonheterosexuality among children of same-sex parents, we found that the higher the maximum rates reported, the less likely those reports were to have been cited in Google Scholar by counts or by annual rate, which may reflect MLSD. However, after several statistical controls, the association for counts became non-significant, while the association for rates became stronger, although the effect sizes were in a moderate (d = .28 or higher) to large range (d, up to .68) by either analysis. Generally, research quality acted as a suppressor variable for MLSD but was significantly related to both counts and rates of citations, indicating that higher quality articles were more likely to have been cited, even controlling for the number of years since first publication. Higher quality articles were slightly more likely to report higher rates of nonheterosexuality among children of same-sex parents. We discuss implications of our findings and suggest future directions of research.
Article
Full-text available
Although widely used in policy debates, the literature on children’s outcomes when raised by same-sex parents mostly relies on small selective samples or samples based on cross-sectional survey data. This has led to a lack of statistical power and the inability to distinguish children born to same-sex parents from children of separated parents. We address these issues by using unique administrative longitudinal data from the Netherlands, which was the first country to legalize same-sex marriage. These data include 2,971 children with same-sex parents (2,786 lesbian couples and 185 gay male couples) and over a million children with different-sex parents followed from birth. The results indicate that children raised by same-sex parents from birth perform better than children raised by different-sex parents in both primary and secondary education. Our findings are robust to use of cousin fixed effects and coarsened exact matching to improve covariate balance and to reduce model dependence. Further analyses using a novel bounding estimator suggest the selection on unobserved characteristics would have to be more than three times higher than the selection on observed characteristics to reduce the positive estimates to zero.
Article
In contrast to earlier studies, several recent ones have claimed that stability rates among same-sex couples are similar to those of different-sex couples. This article reexamines these latest accounts and provides new evidence regarding stability rates using three large, nationally representative datasets from the United States and Canada. Confirming the earliest work, we find that same-sex couples are more likely to break up than different-sex couples. We find that the gap in stability is larger for couples with children, the very group for which concerns about stability are the most important.
Article
Debate persists about whether parental sexual orientation affects children's well-being. This study utilized information from the 2013 to 2015 U.S., population-based National Health Interview Survey to examine associations between parental sexual orientation and children's well-being. Parents reported their children's (aged 4–17 years old, N = 21,103) emotional and mental health difficulties using the short form Strengths and Difficulties Questionnaire (SDQ). Children of bisexual parents had higher SDQ scores than children of heterosexual parents. Adjusting for parental psychological distress (a minority stress indicator) eliminated this difference. Children of lesbian and gay parents did not differ from children of heterosexual parents in emotional and mental health difficulties, yet, the results among children of bisexual parents warrant more research examining the impact of minority stress on families.
Article
In 1996 (Cameron & Cameron) we reported that adult children of homosexuals more frequently reported homosexual desires and sex with their parent(s). Schumm (2015 Schumm, W. R. (2015). Sarantakos's research on same-sex parenting in Australia and New Zealand: Importance, substance, and corroboration with research from the United States. Comprehensive Psychology, 4(16), 1–29. doi:10.2466/17.cp.4.16[CrossRef]) stated he would focus on “the multifaceted research of Sarantakos (1996 Cameron, P., & Cameron, K. (1996). Homosexual parents. Adolescence, 31(124), 757–776.[PubMed], [CSA]) and any research that has corroborated his findings,” but neglected to include ours, though published contemporaneously with Sarantakos.
Article
Full-text available
After lesbian couples have decided to become parents, their family-making journey entails a wide range of encounters with professionals in fertility clinics and/or in maternal and child healthcare services. The article presents the results of an analysis of 96 lesbian mothers' interview talk about such encounters. In their stories and accounts, the interviewees draw on two separate and contradictory interpretative repertoires, the 'just great' repertoire and the 'heteronormative issues' repertoire. Throughout the interviews, the 'just great' repertoire strongly predominates, while the 'heteronormative issues' repertoire is rhetorically minimized. The recurrent accounts of health services as 'just great', and the mitigation of problems, are meaningful in relation to a broader discursive context. In a society where different-sex parents are the norm, the credibility of other kinds of parenthood is at stake. The 'just great' repertoire has a normalizing function for lesbian mothers, while the 'heteronormative issues' repertoire resists normative demands for adaptation. © The Author(s) 2013 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav.
Article
Full-text available
Cigarette smoking has long been a target of public health intervention because it substantially contributes to morbidity and mortality. Individuals in different-sex marriages have lower smoking risk (i.e., prevalence and frequency) than different-sex cohabiters. However, little is known about the smoking risk of individuals in same-sex cohabiting unions. We compare the smoking risk of individuals in different-sex marriages, same-sex cohabiting unions, and different-sex cohabiting unions using pooled cross-sectional data from the 1997–2010 National Health Interview Surveys (N = 168,514). We further examine the role of socioeconomic status (SES) and psychological distress in the relationship between union status and smoking. Estimates from multinomial logistic regression models reveal that same-sex and different-sex cohabiters experience similar smoking risk when compared to one another, and higher smoking risk when compared to the different-sex married. Results suggest that SES and psychological distress factors cannot fully explain smoking differences between the different-sex married and same-sex and different-sex cohabiting groups. Moreover, without same-sex cohabiter’s education advantage, same-sex cohabiters would experience even greater smoking risk relative to the different-sex married. Policy recommendations to reduce smoking disparities among same-sex and different-sex cohabiters are discussed.
Article
Full-text available
A legacy of research finds that marriage is associated with good health. Yet same-sex cohabitors cannot marry in most states in the United States and therefore may not receive the health benefits associated with marriage. We use pooled data from the 1997 to 2009 National Health Interview Surveys to compare the self-rated health of same-sex cohabiting men (n = 1,659) and same-sex cohabiting women (n = 1,634) with that of their different-sex married, different-sex cohabiting, and unpartnered divorced, widowed, and never-married counterparts. Results from logistic regression models show that same-sex cohabitors report poorer health than their different-sex married counterparts at the same levels of socioeconomic status. Additionally, same-sex cohabitors report better health than their different-sex cohabiting and single counterparts, but these differences are fully explained by socioeconomic status. Without their socioeconomic advantages, same-sex cohabitors would report similar health to nonmarried groups. Analyses further reveal important racial-ethnic and gender variations.
Article
It is shown how, in regular parametric problems, the first-order term is removed from the asymptotic bias of maximum likelihood estimates by a suitable modification of the score function. In exponential families with canonical parameterization the effect is to penalize the likelihood by the Jeffreys invariant prior. In binomial logistic models, Poisson log linear models and certain other generalized linear models, the Jeffreys prior penalty function can be imposed in standard regression software using a scheme of iterative adjustments to the data.
Article
Recent legal cases before the Supreme Court of the United States were challenging federal definitions of marriage created by the Defense of Marriage Act and California’s voter approved Proposition 8 which limited marriage to different-sex couples only. Social science literature regarding child well-being was being used within these cases, and the American Sociological Association sought to provide a concise evaluation of the literature through an amicus curiae brief. The authors were tasked in the assistance of this legal brief by reviewing literature regarding the well-being of children raised within same-sex parent families. This article includes our assessment of the literature, focusing on those studies, reviews and books published within the past decade. We conclude that there is a clear consensus in the social science literature indicating that American children living within same-sex parent households fare just, as well as those children residing within different-sex parent households over a wide array of well-being measures: academic performance, cognitive development, social development, psychological health, early sexual activity, and substance abuse. Our assessment of the literature is based on credible and methodologically sound studies that compare well-being outcomes of children residing within same-sex and different-sex parent families. Differences that exist in child well-being are largely due to socioeconomic circumstances and family stability. We discuss challenges and opportunities for new research on the well-being of children in same-sex parent families.
Article
The present study advances research on union status and health by providing a first look at alcohol use differentials among different-sex and same-sex married and cohabiting individuals using nationally representative population-based data (National Health Interview Surveys 1997–2011, N = 181,581). The results showed that both same-sex and different-sex married groups reported lower alcohol use than both same-sex and different-sex cohabiting groups. The results further revealed that same-sex and different-sex married individuals reported similar levels of alcohol use, whereas same-sex and different-sex cohabiting individuals reported similar levels of alcohol use. Drawing on marital advantage and minority stress approaches, the findings suggest that it is cohabitation status—not same-sex status—that is associated with elevated alcohol rates.
Article
Almost all studies of same-sex parenting have concluded there is ``no difference'' in a range of outcome measures for children who live in a household with same-sex parents compared to children living with married opposite-sex parents. Recently, some work based on the U.S. census has suggested otherwise, but those studies have considerable drawbacks. Here, a 20% sample of the 2006 Canada census is used to identify self-reported children living with same-sex parents, and to examine the association of household type with children's high school graduation rates. This large random sample allows for control of parental marital status, distinguishes between gay and lesbian families, and is large enough to evaluate differences in gender between parents and children. Children living with gay and lesbian families in 2006 were about 65% as likely to graduate compared to children living in opposite sex marriage families. Daughters of same-sex parents do considerably worse than sons.