The administrator and the scientist: How the statistical profession has changed

Abstract

The originality of the profession of official statistician comes from its central position between administration and science. Being neither civil servant nor scientist, it combines the traits of both professions, in a different manner according to the eras and countries involved. Four dimensions are proposed to differentiate these modalities. They are linked to the structures of the State, and to the professional profiles of statisticians: centralization of the statistical institution, in both of its forms, administrative and territorial; the location of expertise in the State or outside it; level of skills; mobility and career paths for statisticians. During the 19th century, the administrative dimension dominated. The mathematical tools were poorly developed and often rejected as happened for surveys. The statistical offices were small and not quite legitimate. The legitimacy increased when statisticians managed to associate their work to social or economic questions: public health and hygiene problems (1830–1860), regulation of paid work (1890–1920), macroeconomic Keynesian equilibrium, as of the 1930s. The establishment of the welfare state increased statistical needs and led to the introduction of survey methods.

The “scientification” of the statistical profession came from both transformation of the State and from technical innovations. Four directions can be distinguished. The “representative censuses” are replaced by “sample surveys”. The “statistical counting machines” are replaced by “computers” in the 1950s, allowing the exploitation of administrative sources as complements to the surveys. National accounts, a reflection of the Keynesian macroeconomy, unifies economic statistics. Mathematical statistics leads to the development of econometrics and its modelling.

One way to describe the different aspects of the profession, which correspond to its different tools, is to question the notions of reliability and of precision. These are not purely technical. They are linked to the social use of the statistics presented. The official statistician is confronted with three types of problems, syntax (formalism and internal coherence, the domain of the mathematician), semantical (interpretation, relation to non-statistical language), pragmatic (use of results for action, decision-making and critique). Each of these questions cannot be analyzed without the tension existing between the two dimensions of the statistical profession, those of administrator and scientist.