This study used stochastic frontier analysis to study variations in inefficiency in US hospitals. Cost-inefficiency (i.e., differences between best practice and actual expenses) is assumed to be affected by ownership status, competition, regulatory pressure, and market demand conditions. The level of analysis is the hospital (n = 3,262) and data for 1994 were used. The market was defined as the county in which the hospital was located. A two-stage approach was used in the analysis. In the first stage, translog cost-functions were estimated. Outputs used in the cost function analysis include inpatient discharges, post-admission days, outpatient visits, medical education, and case-mix index. Following Jondrow's technique, inefficiency scores (i.e., the difference between predicted least costs and actual costs) were estimated. Inefficiency estimates were not sensitive to changes in assumptions about the distribution of the error term. In the second stage, the estimated inefficiency scores were used as dependent variables to test hypotheses about the impact of internal and external environmental pressures on cost-inefficiency. Since the distribution of the estimated inefficiency scores was censored, Tobit equations were estimated. The second stage analysis found that measured inefficiency was negatively related with industry concentration (Herfindahl index), public payment policy, and unemployment rate and positively related with for-profit status.