Manipulation and Population Statistics in Nineteenth-Century France and England

Social Research, Summer, 2001 by Libby Schweber

THE status of statistics has shifted radically in the past 30 years. Whereas professionals and politicians once held up statistics as a guarantor of political neutrality and social scientists embraced them as evidence of scientific objectivity, both sets of claims now ring hollow.(1) Or so it seems to the historian or sociologist of science who has followed the past decades of scholarship. Social historians have demonstrated the political intentions behind official statistics (Scott, 1988; Anderson, 1988, 1999), and historians of statistics have revealed the ways that social, political, and professional contexts have shaped the development of specific statistical tools and the interpretation of statistical measures (Brian, 1994; Desrosieres, 1998; Szreter, 1996). One of the most influential studies in this move to historicize statistics was a pathbreaking study by Donald MacKenzie (1978) on the way in which technical debates over the choice of statistical indices were informed by professional and political commitments. More specifically, MacKenzie documented the close articulation between the eugenic and social reform programs of leading statisticians and the "invention" of mathematical statistics in Britain in the first decades of the twentieth century.

My contribution to this special volume on "Numbers" builds on MacKenzie's analysis by comparing his example with a number of nineteenth-century technical-political debates in statistics. The question that interests me concerns the national specificity of MacKenzie's analysis. To what extent was the pattern of fight articulation between technical, cognitive, and political considerations specific to the British statistical community? A second question concerns the relation of modes of discourse and styles of statistical reasoning (where "statistical reasoning" refers to the type of statistical entities and forms of explanation engaged).(2) Finally, I am interested in how this particular pattern of articulation helps to account for Britain's leading role in the development of mathematical statistics at the turn of the century. While a few brief case studies are clearly inadequate to fully address these problems, they can be used to illustrate the questions and to raise a number of suggestions for further research.

The case studies in this paper are part of a larger study comparing the use of population statistics in France and Britain between 1825 and 1885. Whereas most studies in the history of statistics focus on the inventors of new methods and techniques, my own work in this area has focused on practicing statisticians involved in the implementation of new styles of reasoning and the defense of older styles. One of the difficulties with the topic involves the confused and contradictory nature of statisticians' discourse. The introduction of a new way of reasoning is never a simple or linear process. Instead, it involves continual negotiation and reformulation as individuals struggle to make sense of new types of sentences, entities, and forms of explanation. This messiness is particularly evident in the nineteenth-century cases described later, where the absence of formal statistical education and the novelty of the new way of working with numbers meant that participants lacked a common vocabulary or set of standards from which to evaluate knowledge. At the same time it is precisely this openness that reveals the extensive cultural work involved in the introduction of new kinds of statistical abstraction and statistical explanation.

The period between 1825 and 1885 has a peculiar transitional status in the history of statistics. From the perspective of intellectual history, it involved a hiatus between the perfection of probability calculus in the late eighteenth century, most notably in France, and the application of those techniques to social statistics and then physics in the late nineteenth and early twentieth centuries in Britain and Germany (Schneider, 1987; Porter, 1986). From the perspective of institutional and social history, however, the period was marked by an explosion in the production of statistical data, along with the creation and expansion of official statistical bureaus and a conceptual transformation involving the invention of new types of statistical entities, such as population, unemployment and poverty (Hacking, 1990). The novelty of these entities lay in their hidden character. Unlike clans, temporary work, or pauperism, they could be seen only through the medium of statistics. The new project, which is generally ascribed to Adolphe Quetelet, called for the identification of statistical regularities in all aspects of social life.

This search for statistical regularities or laws was associated with a new way of working with numbers and a new set of goals. Whereas previously statistics had been used to describe the population (and as a basis for tax collection), in the new program they were used to infer underlying laws and causes. For the purposes of the discussion, I will refer to the first use as descriptive statistics and the latter as inferential statistics.(3) Far from an easy or obvious activity, the new approach confronted statisticians with the problem of how to manipulate quantitative data and how to interpret the statistical indices they were producing. The problems were particularly telling when it came to the modeling of their own national populations. What acts of aggregation were legitimate? How much abstraction was too much? When did statistical indices cease to correspond to the individuals from whom the original data had been collected? And to what did they correspond?(4)