The literature on prison crowding underscores the potential importance of both state- and facility-level effects on crowding, although empirical research has not assessed these relative effects because of the sole focus on states as units of analysis. This article describes findings from bi-level analyses of crowding across 459 state-operated facilities in 45 states. Findings from cross-sectional models of prison crowding in 1995 and 2000, measured as a ratio of a facility's inmate population to its design capacity, revealed significant differences in levels of crowding based on costs, types of inmates held, and court orders (at the facility level), as well as sentencing policies and the annual number of drug arrests (at the state level). A longitudinal analysis also revealed that differences in crowding levels between 1995 and 2000 coincided with changes in whether facilities increased their design capacity and with rising costs of incarceration. The implications of these findings are discussed. (Contains 8 tables and 3 notes.)