In 2010, Demography published Rosenfeld’s analysis of Census data showing no difference between children living with same-sex couples and children in married families on the only measure of child well-being the Census contains: normal school progress. A new re-evaluation of this study by Prof. Douglas Allen and colleagues was just published in Demography. Rosenfeld reached his conclusions after excluding from the analysis children who were not the “own child” of the respondent (i.e. adopted and stepchildren) and even more oddly, also any family which had moved in the past 5 years.
Allen and colleagues show two things. First, Rosenfeld’s methods yield a conclusion that children with same-sex couples are “no different” from every family form (except divorced men), including family forms with well-known disadvantages, such as never-married mothers. His method is unable to detect in other words any statistically significant difference between children living with single moms, children living with gay couples, and children living in intact married households. But of course in reality children living with same-sex couples who are “no different” on school progress from single moms would be doing radically different from children with married parents. The problem is in the study design, which lacks statistical power to detect real differences.