Research on abusive supervision has predominantly focused on between-person differences in average levels of abusive supervision. Recently, researchers have also demonstrated the importance of considering dynamic, within-person fluctuations in abusive supervision over time. Drawing on uncertainty management theory, I integrate these two streams of research and introduce the concept of “abusive supervision variability”, which captures between- person differences in the stability of abusive supervision over time. To further delineate the consequences of abusive supervision variability, I expand uncertainty management theory by integrating insights from resource depletion theory and affective events theory, respectively, to build and test a mediation chain, dual-process (resource-based and emotion-based) model. Results from a multilevel, experience-sampling field study indicate that abusive supervision variability resulted in greater uncertainty among subordinates, which in turn, depleted subordinate cognitive resources and triggered feeling of anxiety, respectively, which ultimately and jointly, decreased subordinate task performance and organizational citizenship behaviors, but increased counterproductive work behaviors. This study challenges and extends theory and research on abusive supervision by demonstrating that, beyond the intuitive notion that average lower level of abusive supervision is better, consistent experience of abusive supervision can be better for subordinates than sporadic experience of abusive supervision.