How can we better measure how intersecting identities shape health disparities? Evans et al. (2018) show that multilevel models offer an effective, practical way to study health inequalities across multiple overlapping social identities.
Looking at how overlapping identities interact with systems of oppression and privilege is a central focus of health inequality research. Evans et al. present multilevel models as a new alternative for studying these intersections. By splitting the total variation across two distinct levels, this method draws attention to the differences that exist both between separate groups and within the same groups.
To test their approach, the authors used a familiar example from social epidemiology: body mass index (BMI) measurements from 32,788 U.S. adults. Researchers often need to explore many combinations of identities and social dynamics. Here, Evans et al. compared multilevel models to the standard fixed-effects approach across five dimensions of social identity and standing—gender, race or ethnicity, income, education, and age—which combined into a total of 384 unique intersectional groups.
The authors found that the multilevel approach offers several advantages over conventional models, including the ability to handle higher dimensions, adjust for different group sample sizes, keep models simple, and make findings easier to interpret. Furthermore, they discovered that a large portion of the overall variation occurred within the specific groups themselves. This showed that these intersectional categories had low accuracy for sorting individual risk on their own, leaving high differences in risk unexplained within each group. In conclusion, Evans et al. show that multilevel modeling is an innovative and valuable tool for analyzing intersectionality in health inequalities.
Learn more about this article here:
Evans, C. R., Williams, D. R., Onnela, J.-P., & Subramanian, S. V. (2018). A multilevel approach to modeling health inequalities at the intersection of multiple social identities. Social Science & Medicine, 203, 64–73. https://doi.org/10.1016/j.socscimed.2017.11.011