[Form is substance: the construction of national statistical apparatuses (1800-1945)].

The present paper surveys the discussions conducted by official statisticians regarding the ideal structure according to which a national data collection system should have been designed if it was to meet the challenges put up by the various transformation Western countries have undergone since the beginning of the 19th century. Arguments in favour of coordination, centralization, or decentralization have emerged for the first time in 1832 Britain, when the Statistical Bureau of the Board of Trade was created. Up to 1945, this debate went on, the industrial take-off, the economic crises, and the world wars all being occasions for its protagonists to put forward their preferred view. The perspective we take here is original in two respects: on the one hand, instead of confining ourselves to the major statistical systems (those of France, Britain, and the USA), we intend to evoke a large number of cases and, from this comparative standpoint, propose a general account of the drive towards centralization; on the other hand, instead of restraining our-selves to the 19th century, we cover the entire time-frame extending from 1800 to 1945.