Amazingly, one is used to hearing harsh statements about inefficient public services. Nor is it surprising to see public sector performance questioned. What is surprising is that what is meant by performance, and how it is measured, does not seem to matter to either the critics or the advocates of the public sector. The purpose of this paper is to suggest a definition, and a way to measure the performance of the public sector or rather of its main components. Our approach is explicitly rooted in the principles of welfare and production economics. We will proceed in four stages. First of all we present what we call the "performance approach" to the public sector. This concept rests on the principal-agent relation that links a principal, i.e., the State, and an agent, i.e., the person in charge of the public sector unit, and on the definition of performance as the extent to which the agent fulfils the objectives assigned by the principal. The performance is then measured by using the notion of productive efficiency and the "best practice" frontier technique. In the second stage we move to the issue of measuring the performance of some canonical components of the public sector (education, health care and railways transport), assuming that there is no constraint as to data availability. The idea is to disentangle the usual confusion between conceptual and data problems. In the third stage, we move to real world data problems. The question is then that given the available data, does it make sense to assess and measure the performance of such public sector activities. The final stage is to explain performance or rather lack thereof and to look at the contribution of such an exercise for public policy. Finally we argue that when the scope is not components but the entirety of the public sector, one should restrict the performance analysis to the outputs and not relate it to inputs.
