Original title: You cannot decompose injustice: interaction, intersection, random effects and simple means in the evolution of MAIHDA
Juan Merlo · September 2026
Researchers often study health inequalities by looking at social groups formed by overlapping identities, like race and gender. A common statistical method called MAIHDA tries to measure these inequalities. Usually, it tries to separate out a special intersectional effect by mathematically adjusting for the separate identity traits. The author argues this standard approach is flawed. Original ideas of intersectionality say social positions are unique wholes, not math puzzles you can break into parts. To fix this, the author proposes using intact strata, which means keeping each social group whole instead of splitting it apart. The paper introduces Simple-Means MAIHDA, or S-MAIHDA. In this updated approach, researchers estimate each group directly from the people inside it, without relying on statistical shrinkage. Interaction is not removed; instead, it is relocated to show how whole social positions interact with broader structural forces like geography, time, or institutions.
Imagine tasting a soup. You might try to understand it by identifying each ingredient separately — the salt, the garlic, the tomato — but the soup's flavor is not the sum of those ingredients; the same ingredients in different proportions or cooking times taste completely different. In this paper, the author argues that a person's social position — defined by race, gender, and class together — is like the soup: you cannot understand it by decomposing it into separate categories. Each position must be seen as a whole, and the structures around it, like geography or institutions, are like the cooking method, changing the final flavor.
Health researchers use MAIHDA to identify which intersectional groups — defined by race, gender, class, and more — face the worst health outcomes. If the standard approach mischaracterizes intersectionality, the resulting rankings and policy priorities could be misleading. This essay argues for a different way to measure those inequalities, one that treats each social position as a whole rather than a sum of parts.
The abstract doesn't say. It is a conceptual essay, so it offers no empirical validation or comparison of S-MAIHDA against the conventional approach on real data. It also does not discuss potential downsides of keeping strata intact, such as small sample sizes within strata.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
You cannot decompose injustice: interaction, intersection, random effects and simple means in the evolution of MAIHDA — Juan Merlo — SocArXiv — doi:10.31235/osf.io/hn463_v1
Abstract licence: CC BY
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.