Abstract
As a follow-up to our 2016 study, this article presents new findings examining the relationship between same-sex family structure and child health using the 2008–2015 National Health Interview Survey (NHIS). After discussing current NIHS data problems, we examine the relationship between family structure and a broad range of child well-being outcomes, including school days lost, behavior, parent-rated health, emotional difficulties, and activity limitations. We find both similarities (school days lost, behavior, parent-rated health) and differences (emotional difficulties and activity limitations) across our two studies using different survey years, but our overall conclusions are robust. We further discuss the implications of our findings for future research on this topic, including how to account for biological relatedness in a study on child health in same-sex families.
Keywords: Same-sex families, Same-sex marriage, Child well-being, Family structure, National Health Interview Survey
Population-based research on the health and well-being of individuals in same-sex families is a comparatively new field of study on a relatively small portion of the U.S. population. As such, accessing data sources that are nationally representative, are population-based, and have reliable measures of both health and sexual minority status (e.g., identity, same-sex relationship status, behavior) presents a challenge. Study findings in this unique and fast-changing landscape need to be rigorously validated, with many findings requiring post-publication attention as new data issues are discovered (e.g., Chen and Powell 2015; Prickett et al. 2016). This is especially true when examining same-sex families and child well-being because the results of scientific studies are used to inform policy and legal debates on this politically charged topic.
In this spirit, we are happy to return to our recent Demography article after some complications with the National Health Interview Survey (NHIS) have come to light. Herein, we describe NIHS data problems and discuss the implications for our original study. We then report new findings from analyses with new data that examine the same broad range of child well-being outcomes as reported in the original article, demonstrating that our original conclusions mostly hold. Finally, we discuss the implications of our findings for future research on this topic, including how to account for biological relatedness in a study on child health in same-sex families.
The National Health Interview Survey
Scholars have consistently requested data on same-sex families, especially on the children of same-sex married parents (Gates et al. 2011; Manning et al. 2014). This demand culminated in a 2011 Institute of Medicine (IOM) report and a subsequent National Institutes of Health (NIH) call for proposals on sexual minority health (IOM 2011). Six years after this call, there is still a dearth of nationally representative, population-based data sets that include same-sex families. Existing data sources have a number of issues in identifying sexual minorities, and the NHIS is no exception. Still, many scholars have used the NHIS to examine same-sex families, articulating the challenges and rewards of this endeavor (e.g., Denny et al. 2013; Liu et al. 2013; Reczek et al. 2014; Ward et al. 2014). With previously identified data challenges in mind, we took special care in our Demography article to note known data limitations as best as possible. Recent inspection has revealed that many households marked as same-sex couples in four years of data (2004–2007) were mislabeled and are likely different-sex couples. The misidentification was determined via a name-matching process, which may prove problematic given that names in sexual minority communities may take androgynous or nontraditional forms in order to reflect anti-heteronormative performances of sexual or gender identity (Division of Health Interview Statistics 2015). Nonetheless, the process of data cleaning suggests that a number of the same-sex married couples presented in the 2004–2007 years may be mischaracterized.
To assess whether issues in the 2004–2007 data years skewed our original results, we replicate our original analysis excluding the four years of data that have purported invalid measures of same-sex married households. We reanalyze all original outcomes to test whether any of our child well-being findings changed in significance. The inclusion of all original outcome variables is important: as we stated in the original article, family structure may be differentially associated with various dimensions of child well-being. Thus, we replicate our study completely in order to accurately assess our original research question. The methods used in this response are identical to those used in the original study (Reczek et al. 2016) with the exception that we exclude cases before 2008. We also add the 2014–2015 data, which became available in the time since our original analysis was conducted. Please see the original article for a detailed account of our approach (Reczek et al. 2016), including our original sociodemographic covariates that are central in understanding the basic relationship between family structure and child health (Brown 2004; McLanahan and Percheski 2008). Addressing concerns regarding biological relatedness, we also run our models omitting the family relatedness variable (see upcoming discussion).
Results
Our new sample uses data from 2008–2015 and includes 47 same-sex married households, in which there are 72 children (Table 1). For context, we included 125 same-sex married households (n = 125) and 216 children in those households in the original article. (Our original number of same-sex married parent households was 10 below the estimates provided by the NCHS and the commentary because of missing data on 10 of the households.) Our regression results using the 2008–2015 data show both similarities and differences from the original study, but our overall conclusions hold. Our results here focus primarily on same-sex married parent families, but notable results for other groups are also discussed. The final model (Model 2) of each analysis is discussed. Model 1 is also included in the tables, which provides results without controlling for socioeconomic variables, for the sake of comparison with the original study.
Table 1.
Survey-weighted descriptive statistics for child health outcomes by family structure
| Variable | Same-Sex Married | Same-Sex Cohabiting | Different-Sex Married | Different-Sex Cohabiting |
|---|---|---|---|---|
| Parent-Rated Health (ref. = excellent/very good) | ||||
| Good/fair/poor health | ||||
| % | 8.0 | 22.2 | 13.0 | 20.0 |
| (4.3) | (3.2) | (0.2) | (0.6) | |
| N | 72 | 321 | 124,293 | 14,542 |
| Activity Limitations (ref. = no limitations) | ||||
| Has limitations | ||||
| % | 24.3 | 14.3 | 7.1 | 8.9 |
| (5.7) | (2.7) | (0.1) | (0.3) | |
| N | 72 | 321 | 124,191 | 14,522 |
| Lost School Days | ||||
| Mean | 2.3 | 3.3 | 2.7 | 3.0 |
| (0.6) | (0.5) | (0.0) | (0.1) | |
| N | 29 | 119 | 39,429 | 3,906 |
| Emotional Difficulties (ref. = none) | ||||
| Minor | ||||
| % | 14.9 | 24.3 | 12.3 | 19.0 |
| (8.3) | (4.6) | (0.1) | (0.8) | |
| Definite | ||||
| % | 24.1 | 7.5 | 3.2 | 5.8 |
| (10.0) | (3.1) | (0.1) | (0.5) | |
| Severe | ||||
| % | –– | 3.2 | 0.9 | 2.1 |
| (2.1) | (0.0) | (0.3) | ||
| N | 33 | 132 | 42,475 | 4,386 |
| Child Well-behaved (ref. = certainly) | ||||
| Somewhat true | ||||
| % | 21.4 | 25.5 | 17.0 | 26.1 |
| (9.5) | (4.9) | (0.3) | (0.9) | |
| Not true | ||||
| % | 3.3 | 4.5 | 2.2 | 4.9 |
| (3.3) | (2.3) | (0.1) | (0.5) | |
| N | 32 | 112 | 33,347 | 3,578 |
Note: Numbers in parentheses are standard errors.
Source: National Health Interview Surveys, 2008–2015.
Overall, the majority of our results are consistent with those of our original study, as shown in Table 2 for parent-rated health, Table 3 for activity limitations, Table 4 for lost school days, Table 5 for emotional difficulty and Table 6 for behavior problems. We maintain the same order of tables as the original article for ease of comparison. In the new results of Model 2, the basic relationships for child’s parent-rated health, lost school days, and behavior were either similar to our original article or made the original conclusions even stronger. The results regarding parent-rated health (Table 2) are no different in direction or significance from the original analyses. Children of same-sex married parents do not differ from those of different-sex married parents, same-sex cohabiting parents, or different-sex cohabiting parents. The results regarding lost school days (Table 4) are mostly similar to those of the original study; same-sex married families are not significantly different from the different-sex married in child lost school days. Yet, the updated results show fewer significant differences between family types. In the original study, same-sex married parent families reported fewer lost school days on average than same-sex cohabiting parent families. The updated results show no difference between these two family structures, suggesting that the relationship may have been due to misidentified different-sex married families categorized as same-sex married, or due to the reduced sample size. In regard to behavior problems (Table 6), same-sex married parents in the original study were more likely to report behavioral problems than different-sex married families. Our new analyses show no significant difference between same-sex married parent families and other families on behavior problems. Additionally, same-sex cohabiting parents’ children do not differ from other children in the new analysis after socioeconomic covariates are controlled, whereas they reported more severe behavioral problems in the original data. These findings suggest that the observed differences between the children of same-sex married and different-sex married parents may have been a result of data errors rather than an actual trend of disadvantage for same-sex married parent families.
Table 2.
Estimated odds ratios from logistic regression models of parent-rated health on family structure (N = 139,209)
| Model 1 | Model 2 | |||
|---|---|---|---|---|
|
|
||||
| Variable | OR | SE | OR | SE |
| Family Structure (ref. = different-sex married) | ||||
| Same-sex married | 0.68 | 0.40 | 0.89 | 0.52 |
| Same-sex cohabiting | 1.92** | 0.36 | 1.71** | 0.33 |
| Different-sex cohabiting | 1.58*** | 0.07 | 1.25*** | 0.05 |
| Demographic Variables | ||||
| Child female (ref. = child male) | 0.95** | 0.02 | 0.95** | 0.02 |
| Child age | 1.05*** | 0.00 | 1.04*** | 0.00 |
| Child race (ref. = non-Hispanic white) | ||||
| Non-Hispanic black | 1.53** | 0.19 | 1.32* | 0.16 |
| Hispanic white | 2.04*** | 0.09 | 1.42*** | 0.07 |
| Hispanic black | 1.97*** | 0.28 | 1.37* | 0.21 |
| Other | 1.52*** | 0.12 | 1.31** | 0.11 |
| Child foreign-born (ref. = child native-born) | 0.81*** | 0.04 | 0.80*** | 0.04 |
| Parents’ mean age | 0.99*** | 0.00 | 1.00 | 0.00 |
| Parents’ racial composition (ref. = both white) | ||||
| Both black | 1.20 | 0.15 | 1.30* | 0.17 |
| Both other race | 1.09 | 0.10 | 1.41*** | 0.13 |
| Interracial | 0.71*** | 0.04 | 0.86** | 0.05 |
| Parents’ nativity (ref. = both native-born) | ||||
| One parent foreign-born | 1.00 | 0.05 | 1.02 | 0.05 |
| Both parents foreign-born | 1.41*** | 0.07 | 0.96 | 0.05 |
| Family relatedness (ref. = no biological parents) | ||||
| One or more biological parents | 0.57*** | 0.05 | 0.58*** | 0.05 |
| Unspecified parentage | 0.47*** | 0.05 | 0.49*** | 0.05 |
| Region (ref. = Northeast) | ||||
| North Central/Midwest | 1.07 | 0.05 | 1.00 | 0.05 |
| South | 1.03 | 0.05 | 0.96 | 0.04 |
| West | 1.07 | 0.05 | 0.99 | 0.04 |
| Survey Year | 1.01 | 0.01 | 1.01 | 0.01 |
| Socioeconomic Resources | ||||
| Parents’ mean education | 0.89*** | 0.00 | ||
| Parents’ employment (ref. = both employed) | ||||
| One employed | 1.14*** | 0.03 | ||
| Neither employed | 1.54*** | 0.09 | ||
| In poverty | 1.52*** | 0.06 | ||
| Child insured | 1.01 | 0.05 | ||
Note: No significant differences were found between same-sex cohabiting, same-sex married, and different-sex cohabiting groups.
p < .05;
p < .01;
p < .001
Table 3.
Estimated odds ratios from logistic regression models of activity limitations on family structure (N = 139,087)
| Model 1 | Model 2 | |||
|---|---|---|---|---|
|
|
||||
| Variable | OR | SE | OR | SE |
| Family Structure (ref. = different-sex married) | ||||
| Same-sex married | 3.37*** | 1.12 | 3.74***b | 1.28 |
| Same-sex cohabiting | 1.74* | 0.42 | 1.64*a | 0.38 |
| Different-sex cohabiting | 1.37***a | 0.06 | 1.19***a | 0.05 |
| Demographics | ||||
| Child female (ref. = child male) | 0.53*** | 0.01 | 0.52*** | 0.01 |
| Child age | 1.03*** | 0.00 | 1.03*** | 0.00 |
| Child race (ref. = non-Hispanic white) | ||||
| Non-Hispanic black | 1.28* | 0.16 | 1.22 | 0.15 |
| Hispanic white | 1.06 | 0.05 | 0.95 | 0.05 |
| Hispanic black | 1.34 | 0.22 | 1.18 | 0.18 |
| Other | 1.00 | 0.08 | 0.95 | 0.08 |
| Child foreign-born (ref. = child native-born) | 0.84** | 0.06 | 0.88 | 0.06 |
| Parents’ mean age | 1.01** | 0.00 | 1.01*** | 0.00 |
| Parents’ racial composition (ref. = both white) | ||||
| Both black | 0.73* | 0.09 | 0.72** | 0.09 |
| Both other race | 0.87 | 0.09 | 0.93 | 0.09 |
| Interracial | 1.01 | 0.06 | 1.05 | 0.06 |
| Parents’ nativity (ref. = both native-born) | ||||
| One parent foreign-born | 0.87** | 0.05 | 0.88* | 0.05 |
| Both parents foreign-born | 0.53*** | 0.03 | 0.44*** | 0.03 |
| Family relatedness (ref. = no biological parents) | ||||
| One or more biological parents | 0.35*** | 0.03 | 0.37*** | 0.03 |
| Unspecified parentage | 0.31*** | 0.03 | 0.33*** | 0.03 |
| Region (ref. = Northeast) | ||||
| North Central/Midwest | 0.77*** | 0.03 | 0.76*** | 0.03 |
| South | 0.66*** | 0.03 | 0.65*** | 0.03 |
| West | 0.69*** | 0.03 | 0.67*** | 0.03 |
| Survey Year | 1.04*** | 0.01 | 1.04*** | 0.01 |
| Socioeconomic Resources | ||||
| Parents’ mean education | 0.97*** | 0.01 | ||
| Parents’ employment (ref. = both employed) | ||||
| One employed | 1.34*** | 0.04 | ||
| Neither employed | 2.27*** | 0.14 | ||
| In poverty | 1.36*** | 0.06 | ||
| Child insured | 0.63*** | 0.04 | ||
Differences between union statuses, with same-sex married as the reference group, are statistically significant.
Differences between union statuses, with same-sex cohabiting as the reference group, are statistically significant.
p < .05;
p < .01;
p < .001
Table 4.
Estimated regression coefficients from OLS regression models of lost school days on family structure (N = 119,014)
| Model 1 | Model 2 | |||
|---|---|---|---|---|
|
|
||||
| Variable | b | SE | b | SE |
| Family Structure (ref. = different-sex married) | ||||
| Same-sex married | −0.45 | 0.63 | −0.31 | 0.64 |
| Same-sex cohabiting | 0.51 | 0.49 | 0.47 | 0.48 |
| Different-sex cohabiting | 0.29*** | 0.08 | 0.20* | 0.08 |
| Demographics | ||||
| Child female (ref. = child male) | 0.08* | 0.04 | 0.08* | 0.04 |
| Child age | 0.02** | 0.01 | 0.02* | 0.01 |
| Child race (ref. = non-Hispanic white) | ||||
| Non-Hispanic black | 0.59** | 0.18 | −0.64*** | 0.18 |
| Hispanic white | 0.31*** | 0.06 | −0.46*** | 0.07 |
| Hispanic black | −0.53* | 0.20 | −0.67** | 0.20 |
| Other | 0.30** | 0.11 | −0.35** | 0.11 |
| Child foreign-born (ref. = child native-born) | 0.19* | 0.08 | −0.19* | 0.08 |
| Parents’ mean age | 0.01 | 0.00 | 0.00 | 0.00 |
| Parents’ racial composition (ref. = both white) | ||||
| Both black | −0.36 | 0.19 | −0.34 | 0.19 |
| Both other race | 0.24 | 0.13 | −0.15 | 0.13 |
| Interracial | 0.17* | 0.07 | 0.23** | 0.07 |
| Parents’ nativity (ref. = both native-born) | ||||
| One parent foreign-born | 0.26*** | 0.06 | −0.25*** | 0.06 |
| Both parents foreign-born | −0.72*** | 0.06 | −0.87*** | 0.06 |
| Family relatedness (ref. = no biological parents) | ||||
| One or more biological parents | 0.30* | 0.14 | 0.35* | 0.14 |
| Unspecified parentage | 0.42** | 0.15 | 0.48** | 0.15 |
| Region (ref. = Northeast) | ||||
| North Central/Midwest | 0.17** | 0.06 | −0.18** | 0.06 |
| South | 0.10 | 0.06 | −0.12 | 0.06 |
| West | 0.03 | 0.06 | −0.01 | 0.06 |
| Survey Year | 0.07*** | 0.01 | −0.07*** | 0.01 |
| Socioeconomic Resources | ||||
| Parents’ mean education | −0.05*** | 0.01 | ||
| Parents’ employment (ref. = both employed) | ||||
| One employed | 0.32*** | 0.04 | ||
| Neither employed | 0.69*** | 0.11 | ||
| In poverty | 0.02 | 0.07 | ||
| Child insured | −0.10 | 0.08 | ||
| Constant | 3.05*** | 0.21 | 3.53*** | 0.24 |
Note: No significant differences were found between same-sex cohabiting, same-sex married, and different-sex cohabiting groups.
p < .05;
p < .01;
p < .001
Table 5.
Estimated odds ratios from ordinal logistic regression models of emotional difficulties on family structure (N = 122,556)
| Model 1 | Model 2 | |||
|---|---|---|---|---|
|
|
||||
| Variable | OR | SE | OR | SE |
| Family Structure (ref. = different-sex married) | ||||
| Same-sex married | 3.13* | 1.63 | 3.44* | 1.77 |
| Same-sex cohabiting | 2.19** | 0.57 | 2.07** | 0.55 |
| Different-sex cohabiting | 1.79*** | 0.10 | 1.62*** | 0.09 |
| Demographics | ||||
| Child female (ref. = child male) | 0.60*** | 0.02 | 0.60*** | 0.02 |
| Child age | 1.05*** | 0.00 | 1.05*** | 0.00 |
| Child race (ref. = non-Hispanic white) | ||||
| Non-Hispanic black | 1.06 | 0.18 | 1.01 | 0.18 |
| Hispanic white | 0.79*** | 0.04 | 0.72*** | 0.04 |
| Hispanic black | 0.91 | 0.19 | 0.83 | 0.18 |
| Other | 0.91 | 0.08 | 0.88 | 0.08 |
| Child foreign-born (ref. = child native-born) | 0.83* | 0.07 | 0.85* | 0.07 |
| Parents’ mean age | 0.98*** | 0.00 | 0.99*** | 0.00 |
| Parents’ racial composition (ref. = both white) | ||||
| Both black | 0.86 | 0.15 | 0.85 | 0.15 |
| Both other race | 0.81 | 0.10 | 0.86 | 0.10 |
| Interracial | 1.10 | 0.07 | 1.14* | 0.07 |
| Parents’ nativity (ref. = both native-born) | ||||
| One parent foreign-born | 0.88* | 0.05 | 0.89* | 0.05 |
| Both parents foreign-born | 0.61*** | 0.04 | 0.53*** | 0.03 |
| Family relatedness (ref. = no biological parents) | ||||
| One or more biological parents | 0.23*** | 0.02 | 0.24*** | 0.02 |
| Unspecified parentage | 0.19*** | 0.02 | 0.20*** | 0.02 |
| Region (ref. = Northeast) | ||||
| North Central/Midwest | 1.11* | 0.06 | 1.10 | 0.06 |
| South | 1.09 | 0.05 | 1.08 | 0.05 |
| West | 1.08 | 0.06 | 1.06 | 0.06 |
| Survey Year | 1.04*** | 0.01 | 1.04*** | 0.01 |
| Socioeconomic Resources | ||||
| Parents’ mean education | 0.97*** | 0.01 | ||
| Parents’ employment (ref. = both employed) | ||||
| One employed | 1.19*** | 0.04 | ||
| Neither employed | 1.85*** | 0.14 | ||
| In poverty | 1.33*** | 0.07 | ||
| Child insured | 0.84** | 0.05 | ||
| Cut 1 | −0.20 | 0.15 | −0.38* | 0.19 |
| Cut 2 | 1.34*** | 0.16 | 1.17*** | 0.19 |
| Cut 3 | 2.89*** | 0.17 | 2.72*** | 0.20 |
Note: No significant differences were found between same-sex cohabiting, same-sex married, and different-sex cohabiting groups.
p < .05;
p < .01;
p < .001
Table 6.
Estimated odds ratios from ordinal logistic regression models of behavioral problems on family structure (N = 112,692)
| Model 1 | Model 2 | |||
|---|---|---|---|---|
|
|
||||
| Variable | OR | SE | OR | SE |
| Family Structure (ref. = different-sex married) | ||||
| Same-sex married | 1.26 | 0.61 | 1.52 | 0.74 |
| Same-sex cohabiting | 1.71* | 0.46 | 1.58 | 0.45 |
| Different-sex cohabiting | 1.74*** | 0.09 | 1.51*** | 0.08 |
| Demographic Variables | ||||
| Child female (ref. = child male) | 0.67*** | 0.02 | 0.67*** | 0.02 |
| Child age | 0.99** | 0.01 | 0.98*** | 0.01 |
| Child race (ref. = non-Hispanic white) | ||||
| Non-Hispanic black | 1.00 | 0.16 | 0.92 | 0.15 |
| Hispanic white | 1.13* | 0.06 | 0.88* | 0.05 |
| Hispanic black | 1.43 | 0.28 | 1.17 | 0.23 |
| Other | 1.20 | 0.12 | 1.09 | 0.11 |
| Child foreign-born (ref. = child native-born) | 0.92 | 0.07 | 0.92 | 0.08 |
| Parents’ mean age | 0.98*** | 0.00 | 0.99 | 0.00 |
| Parents’ racial composition (ref. = both white) | ||||
| Both black | 0.99 | 0.16 | 1.02 | 0.16 |
| Both other race | 0.66** | 0.08 | 0.78* | 0.10 |
| Interracial | 0.91 | 0.06 | 1.00 | 0.06 |
| Parents’ nativity (ref. = both native-born) | ||||
| One parent foreign-born | 1.04 | 0.06 | 1.06 | 0.06 |
| Both parents foreign-born | 1.03 | 0.06 | 0.78*** | 0.05 |
| Family relatedness (ref. = no biological parents) | ||||
| One or more biological parents | 0.53*** | 0.06 | 0.54*** | 0.06 |
| Unspecified parentage | 0.38*** | 0.05 | 0.39*** | 0.05 |
| Region (ref. = Northeast) | ||||
| North Central/Midwest | 1.11 | 0.07 | 1.07 | 0.07 |
| South | 1.10 | 0.07 | 1.06 | 0.06 |
| West | 1.08 | 0.07 | 1.04 | 0.07 |
| Survey Year | 1.00 | 0.02 | 1.01 | 0.02 |
| Socioeconomic Resources | ||||
| Parents’ mean education | 0.92*** | 0.01 | ||
| Parents’ employment (ref. = both employed) | ||||
| One employed | 1.08* | 0.04 | ||
| Neither employed | 1.43*** | 0.12 | ||
| In poverty | 1.24*** | 0.07 | ||
| Child insured | 1.07 | 0.07 | ||
| Cut 1 | −0.16 | 0.18 | −1.22*** | 0.20 |
| Cut 2 | 2.19*** | 0.18 | 1.14*** | 0.21 |
Note: No significant differences were found between same-sex cohabiting, same-sex married, and different-sex cohabiting groups.
p < .05;
p < .01;
p < .001
The results regarding activity limitations (Table 3) and emotional problems (upcoming Table 5) differ from those in the original article. Children of same-sex married parents in the original article were no more likely than other children to report activity limitations. However, our updated results show that this group is more likely to experience activity limitations than the children of different-sex married parents (odds ratio (OR) = 3.74; p < .001), same-sex cohabiting parents (OR = 2.29 (not shown in table); p < .05), or different-sex cohabiting parents (OR = 3.13 (not shown in table); p < .01). There are no differences between same-sex cohabiting and different-sex cohabiting parents; thus, this does not appear to reflect a same-sex parent disadvantage but rather a trend specific to same-sex married parents. This trend, and its possible association with the adoption of children with special needs by same-sex married parents, is discussed shortly.
The results regarding emotional difficulties (Table 5) also differ from the original study and are mostly consistent with the findings from the commentary. In the original study, same-sex married parent families reported fewer child emotional difficulties than same-sex cohabiting parent families. No other differences were observed for same-sex married parent families in comparison with different-sex families. In the updated analysis, same-sex married parent families report more child emotional difficulties than different-sex married parent families (OR = 3.44; p < .05); however, they do not differ from other families. Notably, in the original study, same-sex cohabiting parents reported significantly more child emotional difficulties than any other family type after we controlled for demographic and socioeconomic factors. In the revised analysis, they report more child emotional difficulties than different-sex married parent families (OR = 2.07; p < .01); they do not differ from other families. The revised analyses do not suggest that children of same-sex parents are uniquely disadvantaged: children of different-sex cohabiting parents also experience higher emotional difficulties than those of different-sex married parents (OR = 1.62; p < .001).
In response to the commentary by Sullins, we conduct the analysis using a dichotomous coding of emotional difficulties (0 = no or minor difficulties; 1 = definite or severe difficulties (referred to as “serious emotional difficulties”)), and find a higher likelihood of same-sex married parents reporting serious emotional difficulties for their children (OR = 6.12; p < .002; 95 % CI = 1.95, 19.23). The direction of findings in the dichotomous model does not differ strongly from our revised findings using ordinal logistic regression, but the estimate is more uncertain. Ultimately, we do not think that the results using an ordinal measure tell a vastly different story from those using a dichotomous measure, especially given the high uncertainty of the dichotomous estimates. Ordinal measures more accurately reflect the responses of parents in the study. And as we discuss shortly, emotional difficulties may be a particularly difficult variable to understand without a reliable parental relatedness variable due to the selection effects of adoption.
In summary, these findings on multiple measures of well-being are an update in what we hope will be a growing body of research on same-sex marriage and child well-being. We find both similarities (school days lost, behavior, parent-rated health) and differences (emotional difficulties and activity limitations) across our two studies using different survey years, but our overall conclusions are robust. Although space precludes us from going in depth as to why we find that same-sex married families have children with more activity limitations and emotional difficulties than the different-sex married, we provided an in-depth account of the various theoretical reasons in our original article. In short, perhaps minority stress caused by institutional and social discrimination for being in a same-sex family is related to child emotional and behavioral problems. Same-sex disadvantages may also reflect selection of adopted or special needs children into same-sex married families, who are more likely to adopt than other family forms (discussed later). Additional research should examine these potential mechanisms.
Parental Biology in the Same-Sex Family Context
The use of the biological status variable is something that we, and the Demography reviewers and editors, discussed at length and decided to include in the original article for several reasons. This variable is used consistently in research on different-sex family structure and child health to examine whether a child is biologically related to both parents in a household: it is seen as a proxy for family instability (Manning and Lamb 2003). Clearly, this becomes more complicated when discussing same-sex families because both parents cannot be biologically related to a child. Because of the collinearity problem between biological relatedness and same-sex parents, we combined two biological parents as one group; the resulting categories are (1) one or two biological parents, (2) no biological parents, and (3) unspecified. Thus, we are merely testing whether a child is biologically related to at least one parent for the entire sample of same- and different-sex families, as a proxy for whether the child is adopted, a foster child, a child from another circumstance, or a biological descendent from one of the same-sex partners. We noticed and interpreted some differences in the biological relatedness variable used in the commentary, which used the categories (1) one biological parent, (2) two biological parents, and (3) no biological parents—this creates the problem with biological status that our variable attempts to bypass. We also note an error in our regression tables labeling for the reference group of biological relatedness, which may have been the cause for confusion on this variable. The reference group should be “no biological parents” instead of “two biological parents” in Tables 2–6 in the original paper. We apologize for this error, although the descriptive table and text are correct.
We hope that this conversation can inform future research with clearer biological and adoptive data. Indeed, our additional analyses omitting family relatedness did not significantly alter the results of our models (Table 7). Importantly, our effort to control biological relatedness is only an initial attempt to tease out this issue; we understand the measure of this variable is limited in the NHIS data. For example, one major issue with the biological relatedness variable is that the NHIS does not always distinguish between adoptive or biological parents. In 2013, the NHIS switched coding for child’s relatedness to parents to be “biological or adopted,” thus making distinctions between the two impossible.
Table 7.
Estimated regression coefficients from final regression models without family relatedness
| Parent-Rated Health | Activity Limitations | Lost School Days | Emotional Difficulties | Behavioral Problems | |
|---|---|---|---|---|---|
| Family Structure | Odds Ratio | Odds Ratio | b | Odds Ratio | Odds Ratio |
| Different-Sex Married (ref.) | 1.00b | 1.00ab | 1.00 | 1.00ab | 1.00 |
| Same-Sex Married | 0.92 | 3.99*** | −0.49 | 4.16* | 1.71 |
| (0.53) | (1.33) | (0.61) | (2.29) | (0.91) | |
| Same-Sex Cohabiting | 1.76** | 1.81** | 0.49 | 2.34** | 1.61 |
| (0.34) | (0.41) | (0.49) | (0.59) | (0.45) | |
| Different-Sex Cohabiting | 1.26*** | 1.21***a | 0.29*** | 1.64*** | 1.51*** |
| (0.06) | (0.05) | (0.08) | (0.09) | (0.08) |
Notes: Numbers in parentheses are standard errors. The following variables were controlled in the models but suppressed in the tables: child’s age, gender, race, nativity status, and insurance status; parents’ mean age, racial composition, nativity status, mean education, and employment status; family relatedness; region; and survey year family poverty status.
Differences between union statuses, with same-sex married as the reference group, are statistically significant.
Differences between union statuses, with same-sex cohabiting as the reference group, are statistically significant.
p < .05;
p < .01;
p < .001
This likely affected our updated results. For activity limitations, 55 (of 72) children of same-sex married parents; 127 (of 321) children of same-sex cohabiting parents; 44,251 (of 124,172) children of different-sex married parents; and 5,616 (of 14,522) children of different-sex cohabiting parents are “unspecified” in the new full file. Across all our dependent variables, the percentage of same-sex married parent children with “unspecified” relatedness varies from 72 % to 76 %. Because of the large “unspecified” category that preempts a clear separation between adopted children and biological children, it is possible that trends in adoption are not properly controlled in the new analysis. For example, according to a recent analysis (Gates 2013:3), “13 % of same-sex couples have an adopted child compared to just 3 % of different-sex couples.” Same-sex parents are more likely to adopt children with special needs or disabilities (Lewin 2009:53), with approximately 50 % of same-sex adoptive parents choosing special needs children (Brodzinsky and Pertman 2011). These higher rates of adoption likely explain some of the disadvantages in emotional difficulties and activity limitations faced by children of same-sex married parents. Unfortunately, without improved data that distinguish between biological and adopted children, we are unable to explore this further.
The Future of Research on Same-Sex Family Health
We appreciate the opportunity to clarify, update, and offer additional commentary on our study. We echo the long-standing call for better data sources in order to more completely understand the basic relationship between same-sex family structure and child well-being. Future data should include not only a wide range of health measures for minor children but also better measures of adoptive and biological status as well as data on the quality of the parent-child tie, social support processes, and sexual minority stress and discrimination. Data on sexual identity, as well as on nonmarried, noncohabiting single parents across sexual identity, will also facilitate better population science on this topic. The NIHS added a sexual orientation measure in 2013 (see Reczek et al. 2017), and this will help uncover more inconsistencies as well as advance the field. Perhaps most importantly, high-quality, nationally representative longitudinal data will allow us to understand both causality in the basic relationship between family structure and child health and the theoretical mechanisms that underlie this basic relationship. Research on same-sex families and child health significantly lags behind research on different-sex families and child health, and new data will help build a clear consensus at the population level. We sincerely hope this conversation is generative toward the collection of new data and research on same-sex families and well-being in the years to come.
References
- Brodzinsky DM, Pertman A, editors. Adoption by lesbians and gay men: A new dimension in family diversity. New York, NY: Oxford University Press; 2011. [Google Scholar]
- Brown SL. Family structure and child well-being: The significance of parental cohabitation. Journal of Marriage and Family. 2004;66:351–367. [Google Scholar]
- Cheng S, Powell B. Measurement, methods, and divergent patterns: Reassessing the effects of same-sex parents. Social Science Research. 2015;52:615–626. doi: 10.1016/j.ssresearch.2015.04.005. [DOI] [PubMed] [Google Scholar]
- Denney JT, Gorman BK, Barrera CB. Families, resources, and adult health: Where do sexual minorities fit? Journal of Health and Social Behavior. 2013;54:46–63. doi: 10.1177/0022146512469629. [DOI] [PubMed] [Google Scholar]
- Division of Health Interview Statistics, National Center for Health Statistics. Changes to data editing procedures and the impact on identifying same-sex married couples: 2004–2007 National Health Interview Survey. Hyattsville, MD: U.S. Department of Health and Human Services Centers for Disease Control and Prevention; 2015. Apr, Retrieved from https://www.cdc.gov/nchs/nhis/nhis_2007_data_release.htm. [Google Scholar]
- Gates GJ. Family formation and raising children among same-sex couples (NCFR report FF51) St. Paul, MN: National Council on Family Relations; 2011. Retrieved from http://williamsinstitute.law.ucla.edu/wp-content/uploads/Gates-Badgett-NCFR-LGBT-Families-December-2011.pdf. [Google Scholar]
- Gates GJ. LGBT parenting in the United States (The Williams Institute research brief) Los Angeles: University of California Los Angeles School of Law; 2013. Feb, Retrieved from http://williamsinstitute.law.ucla.edu/wp-content/uploads/LGBT-Parenting.pdf. [Google Scholar]
- Institute of Medicine of the National Academies Committee on Lesbian, Gay, Bisexual, and Transgender Health Issues and Research Gaps and Opportunities. The health of lesbian, gay, bisexual, and transgender people: Building a foundation for better understanding. Washington, DC: National Academies Press; 2011. [PubMed] [Google Scholar]
- Lewin E. Gay fatherhood: Narratives of family and citizenship in America. Chicago, IL: University of Chicago Press; 2009. [Google Scholar]
- Liu H, Reczek C, Brown D. Same-sex cohabitors and health: The role of gender, race-ethnicity, and socioeconomic status. Journal of Health and Social Behavior. 2013;54:25–45. doi: 10.1177/0022146512468280. [DOI] [PubMed] [Google Scholar]
- Manning WD, Fettro MN, Lamidi E. Child well-being in same-sex parent families: Review of research prepared for American Sociological Association Amicus Brief. Population Research and Policy Review. 2014;33:485–502. doi: 10.1007/s11113-014-9329-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Manning WD, Lamb KA. Adolescent well-being in cohabiting, married, and single-parent families. Journal of Marriage and Family. 2003;65:876–893. [Google Scholar]
- McLanahan S, Percheski C. Family structure and the reproduction of inequalities. Annual Review of Sociology. 2008;34:257–276. [Google Scholar]
- Prickett KC, Martin-Storey A, Crosnoe R. An update on identifying same-sex couples in the American Time Use Survey. Demography. 2016;53:2121–2122. doi: 10.1007/s13524-016-0523-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Reczek C, Liu H, Spiker R. A population-based study of alcohol use in same-sex and different-sex unions. Journal of Marriage and Family. 2014;76:557–579. doi: 10.1111/jomf.12113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Reczek C, Liu H, Spiker R. Self-rated health at the intersection of sexual identity and union status. Social Science Research. 2017;63:242–252. doi: 10.1016/j.ssresearch.2016.09.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Reczek C, Spiker R, Liu H, Crosnoe R. Family structure and child health: Does the sex composition of parents matter? Demography. 2016;53:1605–1630. doi: 10.1007/s13524-016-0501-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ward BW, Dahlhamer JM, Galinsky AM, Joestl SS. National Health Statistics Reports No. 77. Hyattsville, MD: National Center for Health Statistics; 2014. Sexual orientation and health among U.S. adults: National Health Interview Survey, 2013. Retrieved from https://www.cdc.gov/nchs/data/nhsr/nhsr077.pdf. [PubMed] [Google Scholar]