Abstract
This paper reports the results of an empirical comparison of three statistical cost frontier models using data from 49 Spanish public hospitals. The models estimated include the deterministic cost frontier, a cross-section stochastic cost frontier (SCF) in which inefficiency is assumed to follow a half-normal distribution and a panel-data SCF in which inefficiency is assumed to remain constant over time. A non-frontier model is also estimated. The paper compares the estimates of average inefficiency obtained from the frontier models and the implied rankings of hospitals in terms of their 'costliness' (actual cost as a proportion of 'expected' cost)