Measuring technical efficiency in the National Health Service: a stochastic frontier analysis

This paper reports the results of an investigation into the technical efficiency (i.e. the ability to convert inputs into outputs) the NHS hospital sector. The method employed is the ‘stochastic frontier production function’. In contrast to the approach adopted by Feldstein in his ‘Economic Analysis for Health Service Efficiency’, the stochastic frontier approach recognises that a hospital’s failure to produce exactly what would be expected of it on the basis of the parameters of its production function may be due not only to technical inefficiency but also to random influences outside its control (e.g. viruses). The model is estimated on data from 193 maternity hospitals for the financial year 1971/72. Various special cases of a transcendental logarithmic frontier production function are estimated, including a Cobb-Douglas function; all are estimated under the assumption that the ‘error’ term reflecting inefficiency has a half-normal distribution. Surprisingly, in none of the models estimated was there any evidence of technical inefficiency. Interpreted literally, therefore, all the hospitals in the sample were operating at 100% technical efficiency, obtaining maximum ‘output’ (deliveries) from their bundle of inputs.