This is not the first time the NHIS data have been used to analyze same-sex households and child health. A manuscript presented at the 2014 annual meeting of the Population Association of America assessed the same data. Curiously, that manuscript overlooked all emotional health outcomes. Instead, the authors inquired only into a solitary, parent-reported measure of their “perception of the child’s overall health,” a physical well-being proxy that varies only modestly across household types. Hence, the authors readily concluded “no differences.”

I’m not surprised.

This juxtaposition provides a window into the state of the social science of same-sex households with children. Null findings are preferred—and arguably sought—by most scholars and journal editors. Indeed, study results seem to vary by author, not by dataset. It is largely a different approach to the presentation of data that distinguishes those population-based studies hailed by many as proof of “no differences” from those studies denounced by the same people as “junk science.”

In fact, population-based surveys of same-sex households with children all tend to reveal the same thing, regardless of the data source. It’s a testimony to the virtues of random sampling and the vices of relying on nonrandom samples, which Sullins argues—in another published study—fosters “a strong bias resulting in false positive outcomes . . . in recruited samples of same-sex parents.” He’s right. Published research employing the New Family Structures Study (NFSS), the ECLS (Early Childhood Longitudinal Study), the US Census (ACS), the Canadian Census, and now the NHIS all reveal a comparable basic narrative, namely, that children who grow up with a married mother and father fare best at face value.