Although empirical evidence has accumulated showing that abusive supervision has devastating effects on subordinates' work attitudes and outcomes, knowledge about how such behavior impacts supervisors who exhibit it is limited. Drawing upon conservation of resources theory, we develop and test a model that specifies how and when engaging in abusive supervisory behavior has immediate benefits for supervisors. Via two experiments and a multi-wave diary study across 10 consecutive workdays, we found that engaging in abusive supervisory behavior was associated with improved recovery level. Moreover, abusive supervisory behavior had a positive indirect effect on work engagement through recovery level. Interestingly, supplemental analyses suggested that these beneficial effects were short-lived because, over longer periods of time (i.e., one week and beyond), abusive supervisory behavior were negatively related to supervisors' recovery level and engagement. The strength of these short-lived beneficial effects was also bound by personal and contextual factors. Empathic concern--a personal factor--and job demands--a contextual factor--moderated the observed effects. Specifically, supervisors with high empathic concern or low job demands experienced fewer benefits after engaging in abusive supervisory behavior. We discuss the theoretical and practical implications of these findings, and propose future research directions.