This paper presents results from a systematic review and meta-analysis of studies of the predictive efficacy of the Historical, Clinical, and Risk Management-20 (HCR-20) for aggressive behavior in residential psychiatric facilities. Variations in efficacy were investigated based on aggression-type, HCR-20 scale used, and as moderated by clinical, demographic and methodological variables. Comprehensive terms were used to search seven electronic databases between January 1995 and August 2012. Additional papers were located by examining references lists and hand-searching. Twenty non-overlapping studies involving 2067 participants were identified. Few (n = 4) studies reported methodology and results sufficiently to ensure a transparently low risk of bias. The summary judgment (dw = 1.166) had the largest mean effect size for prediction of any inpatient aggression. HCR-20 had best predictive efficacy among samples containing higher proportions of patients with schizophrenia, women, Caucasians, and those with greater risk of bias. Predictive efficacy was reduced in studies containing higher proportions of patients with personality disorder. HCR-20 is a significant predictor of aggression in residential psychiatric facilities but does not appear to have equal efficacy across groups. Future research should aim to verify current findings using more heterogeneous samples and should report methodology with greater rigor.