A serious criticism of the ad hoc UN criminal tribunals is that in pursuing victor’s justice, they systematically discriminated against certain identity groups in issuing decisions on convictions and prison sentences (Scharf 1997; Nizich 2001). We situate this criticism in the context of theories of international relations and international law, and then evaluate it empirically on the complete public records of all ICTY-IRMCT cases, yielding an abundance of defendant-level correlates of trial outcomes. The legal literature’s approach to testing hypotheses of judicial bias has been to present bivariate cross-tabulations of ethnicity and trial outcomes among indictees (Ford 2013; Vajda et al 2019). We explain how sample censoring and confounding bias this approach, and we introduce an alternative recently applied in the study of racial bias in police stops (Knox, Lowe, and Mummolo 2020). We present nonparametric bounds on the treatment effects of ethnicity-related variables on trial outcomes under all plausible assumptions about prosecutorial bias, while using new data to adjust for relevant confounds. Whereas naive analyses of our data suggest that the ICTY-IRMCT trial chambers treated all defendants equally, our partially identified estimates suggest an important exception: Bosnian Serbs. We show that, under reasonable assumptions of prosecutorial “bias” arising from practical constraints on arrest and surrenders from 1995-present, a defendant’s allegiance to Bosnian Serb interests raised their probability of conviction and increased their sentence duration.