Prior research into factors predicting arrest in domestic violence cases is limited in three regards: (a) no examination of whether mandatory arrest policies are associated with increased risk of arrest across multiple jurisdictions; (b) little consideration of whether police organizational characteristics influence arrest in such cases; and (c) little examination of whether mandatory arrest policies moderate the influence of extralegal characteristics on arrest risks. Using year 2000 data from the National Incident-Based Reporting System and the Law Enforcement Management and Administrative Statistics, a nonlinear hierarchical modeling procedure was employed to evaluate the predictive utility of situational and organizational factors associated with arrest probability in domestic violence cases. The results suggest that, in domestic violence cases, mandatory arrest policies predict arrest and modestly reduce the significance of race as a predictor of arrest. The results are interpreted as supporting a constrained rational model of policing. (Contains 1 figure, 3 tables and 11 notes.)