LAUSR

I would like to thank Schwartz [1] for insightful questions about our paper [2] on whether and how quantitative approaches could be useful for understanding intersectionality, and I would also like to thank the Editor for providing an opportunity to respond. On many points, we agree. The additive joint disparity we considered is a policy-relevant measure that can be used to track the health of multiply marginalized populations, without any reference to interactions, and certainly this focus is consistent with intersectionality. Schwartz’s emphasis that, when available, social theory should guide an analysis is an excellent one and is why we provided an extensive, though not exhaustive, review of additive interaction measures and modeling strategies. It would indeed be interesting to evaluate whether tests for synergism in the sufficient-cause framework [3, 4] could be mapped to concepts of intersectionality. Many of Schwartz’s questions revolve around the causal status of race, a topic which has been considerably debated in the statistics, social science, and more recently the epidemiology literature [5–11]. Contrary to Schwarz’s assumption, our study did not consider race as a cause, nor did we consider potential outcomes for race. Early on, we made clear: ‘‘In attributing disparities to race, we mean the historical legacy of racism in the United States (US) taking shape through various means, including slavery, Jim Crow, and segregation [12]. This legacy involves discrimination which may be intentional or the result of policies, laws or practices that systematically disadvantage blacks and shape economic opportunities [13] (and also) indirect processes whereby blacks are more vulnerable to economic or political shifts because of this legacy’’ [14]. The reality of class in the US can reinforce this legacy by mediating access to quality neighborhoods, housing, and education. It is also important to remember that the joint decomposition originated long before potential outcomes was used to infer causal inference in epidemiology [15, 16], before any entrenched perspectives about restricting studies to welldefined interventions arose, so its use does not imply that paradigm. Our use of the joint decomposition focuses on describing and understanding how outcomes are patterned for a multiply marginalized group as compared to a nonmarginalized group. It is not equipped to speak of the action of racism or socioeconomic disadvantage or their intersection at the person level; its focus is on understanding patterns among the population. In what follows, I will explain how, in our descriptive implementation, the joint decomposition relates to certain features of intersectionality. The central point is that even though potential outcomes of race/SES were not considered, the notion of a disparity itself is inherently counterfactual, and this property helped us draw insights about disparities across multiple axes. This property is also what would encourage one to identify targets to reduce those disparities. Accordingly, I will demonstrate how a mediation analysis for the joint disparity and its decomposition can support this aim. I also outline how to repurpose existing software for mediation analysis to accomplish this aim, and also present non-parametric formulae that can be This comment refers to the article available at doi:10.1007/s00127016-1334-0.