Abstract
Racial disparities in healthcare expenditures are well-documented, yet theunderlying drivers remain complex and require further investigation. This studydevelops a framework for decomposing such disparities through shifts in thedistributions of mediating variables, rather than treating race itself as amanipulable exposure. We define disparities as differences incovariate-adjusted outcome distributions across racial groups, and decomposethe total disparity into two components: one attributable to differences inmediator distributions, and another residual component that would remain evenafter equalizing these distributions. Using data from the Medical ExpendituresPanel Survey, we examine the extent to which expenditure disparities wouldpersist or be reduced if mediators such as socioeconomic status, insuranceaccess, health behaviors, or health status were equalized across racial groups.To ensure valid inference, we derive asymptotically linear estimators based oninfluence-function techniques and flexible machine learning tools, includingsuper learners and a two-part model designed for the zero-inflated,right-skewed nature of expenditure data.