Disparities across social categories such as race, gender, and class are central in social stratification. The complexity of these constructs, however, hinders their placement within a causal framework. On one hand, it is difficult to imagine a manipulation to alter the category to which one is assigned. On the other hand, categories themselves may be mutable across time and place as a result of social forces such as government definitions of racial categories. This paper advances gap-closing

more »

... timands that define precise causal research goals without reifying the definitions of social categories or appealing to a hypothetical world in which one's categorization were different. Instead, a gap-closing estimand directs attention to a manipulable treatment variable and asks a causal question: what gap across categories would persist under a local intervention to equalize the treatment? The proposal extends related work from epidemiology in three ways. First, I clarify that the hypothetical intervention is local rather than global in nature; there is no appeal to simultaneously equalize the treatments of the entire population. Second, I formalize equalization at a single treatment value or at a stochastic rule for treatment assignment. Third, I connect these estimands to doubly-robust estimators that combine treatment and outcome modeling. I illustrate with an example about the gap in pay by class origins under an intervention to equalize occupational class destinations. The paper concludes with implications for practice: gap-closing estimands provide tools for the rigorous study of inequality across social categories that could inform policies to close gaps.