Heterogeneity of Error Variance and the Assessment of Moderating Effects of Categorical Variables: A Conceptual Review

Abstract

Homogeneity of within-subgroup error (residual) variance is a necessary statistical assumption for the appropriate use of moderated multiple regression (MMR) for assessing the effects of categorical moderator variables (e.g., ethnicity, gender). We provide a conceptual review of the homogeneity of error variance assumption in the context of MMR analyses. First, we clarify issues pertaining to the violation of the homogeneity of error variance assumption and differentiate it from the homoscedasticity assumption. Second, we delineate the implications of violating the homogeneity of error variance assumption for organizational theory and practice. Finally, we critically review solutions recently proposed to mitigate the detrimental effects of violating the homogeneity of error variance assumption on conclusions regarding the effects of categorical moderator variables.