When we come finally to the institutional factors in the producer perspective, we arrive at the public-choice school. The citizens get the public sector that the politicians (Downs) or the bureaucrats (Niskanen) conceive of and put on the market.49
Discarding or de-emphasizing the budget-maximization hypothesis does not leave one by any means empty-handed when one tries to account for the increase of the public sector in the United States and Western Europe.
To follow the debate concerning bureaucrats as budget-maximizers is to listen to a lively and intelligent, at times even brilliant, conversation in which insights and hypotheses cross swords while the will to test the notion systematically is less well developed. In this theoretically top- heavy debate, with each new model more elegant than the last, the hypothesis lives on as if it stated something about reality, as if it were confirmed through observation and not merely the expression of an obstinate postulate. The aura of legend has arisen around the budget- maximization hypothesis and we become mesmerized into believing in its vitality by being confronted with it everywhere. On the popular television programme 'Yes Minister', Sir Humphrey Appleby says, 'The Civil Service does not make profits or losses. Ergo, we measure success by the size of our staff and budget. By definition, a big department is more successful than a small one . . . this simple proposition is the basis of our whole system.'50
No matter how such quips may tickle our prejudices and strengthen our impressionistic experiences, a scientific description must be founded on more systematic work. In this respect one is in the first place struck by the manifoldness and complexity inherent in the causal explanations offered. To trace the growth of the public sector to a single cause such as the bureaucracy's selfish interest in maximum budgets is, as the survey of the debate has shown, clearly to over- simplify matters. A number of conditions of the most various nature seem to be at work. At the same time, the relations between these conditions are themselves complicated. Simply to run Tarschys's coarse classification through a regression analysis in the hope of determining the explanatory power of the separate factors would not be worth the effort, since the factors are themselves so different from