This dissertation develops and empirically tests a theory of interaction between the federal appellate courts and the bureaucracy with regard to bureaucratic prosecution. Modeling the bureaucracy as a forward-looking and risk-averse institution and assuming that there is no uncertainty at the district court level, I posit that institutional uncertainty created by appellate courts' random assignment of judges and cases affects the likelihood of bureaucratic prosecution. Given that the decision from a specific panel of a circuit court can be estimated by its median judge's policy position and that the bureaucracy does not know which panel will hear the case, there exists institutional uncertainty at the appellate court level in terms of ideological differences among panels represented by their median judges. I contend that increasing ideological heterogeneity within an appellate court measured by its ideological variance among judges increases institutional uncertainty with respect to the bureaucracy's policy position, which in turn discourages bureaucratic prosecution. My examination of the Antitrust Division's prosecution record from 1950 to 1994 demonstrates that ideological variance within the federal circuit courts has a significant impact on the likelihood of prosecution by the agency. The Antitrust Division is less likely to prosecute when facing a circuit court with large ideological variance among judges. Studies of judicial decision-making and judicial control of the bureaucracy have not fully examined the implication of appellate courts' institutional practice of randomly assigning judges and cases. The development of ideological variance among circuit judges, in this project, as a measure of the institutional uncertainty created by the random assignment process suggests that the courts' unique institutional practice can now be fully incorporated into future studies of the interaction between the judiciary and the bureaucracy.