Introducing the multilevel analysis of individual heterogeneity and discriminatory accuracy (MAIHDA) to quantitatively examine intersectional workplace inequities

Introduction Calls to diversify research methodology in organizational intersectionality research are growing to better capture intersectional workplace inequities. This study introduces Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy (MAIHDA) as an advanced quantitative method for intersectionality research. Methods Using Collaborative on Academic Careers in Higher Education (COACHE) faculty job satisfaction survey data from a research-intensive university, we examine satisfaction with institutional leadership across four nested models incorporating up to nine social identities and positionalities. We compare MAIHDA with a saturated unregularized fixed-effects approach (FEA) and ridge-, LASSO-, and elastic-net-regularized interaction models using repeated cross-validation. Results As model dimensionality and sparsity increased, MAIHDA showed more stable predictive performance than the saturated FEA, while regularized approaches performed similarly to MAIHDA. Repeated half-sample analyses showed that some global and stratum-specific MAIHDA findings were sensitive to sample reduction and composition. Simulations similarly demonstrated advantages of MAIHDA over the saturated FEA under the sparse, high-dimensional conditions examined, while identifying specific non-additive strata remained challenging. Discussion These findings demonstrate both the potential and limitations of MAIHDA for organizational intersectionality research and advance critical quantitative inquiry in organizational studies.