The influence of socioeconomic factors on bureaucratic structures was examined first, using the following variables for juvenile corectional organizations: stability of leadership, structural complexity, structural stability, centralization, agency autonomy, fiscal control complexity, level of staff experience, access to the Governor, number of employees, number of offenders, professionalism, and unionization. The agency attributes most responsive to socioeconomic pressures tended to be structural, indicating that agencies become larger, more complex, and formal as the number of offenders they handle increases. Socioeconomic factors, such as income or urbanization, were of little help in predicting staff opinions on key issues. Depencdent policy variables included in the study were State per capita expenditures on juvenile corrections, per offender expenditures for incarcerated juveniles, and deinstitutionalization levels. Analysis showed a strong relationship between staff unionization and per offender expenditures but no ties between professionalism and spending. This implies that in high expenditure States, money is being spent on uniformed and maintenance personnel rather than on professional staff. Both the agency director's ideology and agency autonomy had substantial, interacting effects on policy. Director ideology was relatively unimportant when autonomy was low, but its impact was great when the corrections agency enjoyed substantial decisionmaking powers. A comparison of these findings with other studies suggests that director ideology is most significant when the bureaucracy involved in small, and policy is politically salient within the agency and can be swiftly implemented. The relationship between bureaucratic characteristics and programs is complex and varies among different types of agencies. For example, directors of correctional facilities often play a major role in policymaking because other public officials have few ideas of their own and cannot envision the consequences of particular programs. Future research should use large samples and include all potentially interacting variables. Eight notes are provided.