Technology and Culture

America by the Numbers: Quantification, Democracy, and the Birth of National Statistics By Emmanuel Didier. Translated by Priya Vari Sen. Cambridge, MA: MIT Press, 2020. Pp. 432.

The story of probability theory in the eighteenth and nineteenth centuries—its roots in insurance, law, and statecraft, as well as the subsequent epistemological resistance to it—is now well-established through classic works like Daston's Classical Probability in the Enlightenment (Princeton University Press, 1988), Hacking's Taming of Chance (Cambridge University Press, 1990), and Porter's The Rise of Statistical Thinking (Princeton University Press, 1986). With this book, Didier follows the continuing probabilistic revolution into the early twentieth century in a work that beautifully complements Sarah Igo's Averaged American (Harvard University Press, 2007). Igo established the ways that probability and statistics contributed in this same period to the building of selves in a mass society. Didier's parallel study casts attention on the making of political aggregates at the dawn of the welfare state. Didier attends to the political uses of statistical theory in the tradition of Alain Desrosieres' The Politics of Large Numbers (Harvard University Press, 1998). A new generation of scholars is now extending this history into the mid- and late twentieth century—scholars like Theodora Dryer, Arunabh Ghosh, Chris Phillips, and Alma Steingart. Didier's book is an essential piece in this literature.

This book was originally published in 2009 in French with a title that translates roughly to "What Does America Consist of?" That title captures the book's theoretical core. Didier focuses on the production of two aggregates: crop estimates and unemployment statistics. He argues that in the early twentieth century, America (as depicted in these aggregates) consisted of an assortment of judgments made through distributed, participatory statistical labor. But the introduction of a new technology—statistical sampling—instead turned America into a collection of "sample frames" and "representative" respondents (or non-respondents) whose data could be extracted by professional statisticians serving technocrats. [End Page 608]

Didier begins with the material practices used by Department of Agriculture statisticians in the 1910s and 20s. A statistician in each state enlisted volunteer farmers—"a sort of parliament of the entire agricultural territory"—chosen to represent their regions through reports on local conditions, which the statistician collected and edited, drawing on his own experiences, interviews, and observations from around the state and from past years' reports, shaping from this mass of "plasma" a smoothed map, a weighted chart of the state's estimated yields for each important crop (p. 135). A federal board then released regular estimates that set an informational standard used by farmers, speculators, politicians, and everyone else. Didier traces out this example in depth, using it to illustrate how an aggregate should be seen as a "consolidation" from diverse materials. Readers interested in where these practices came from should read Jamie Pietruska's Looking Forward (Chicago, 2017), published in the intervening decade since Didier's French original.

Having presented this consolidated aggregate, America, Didier traces what he calls its deliquescence. The Great Depression and state responses to it sidelined crop estimates. Alternative aggregates began forming within the crop reporting system. In 1921, Agricultural Department statisticians started experimenting with techniques for gathering objective data about specific farms, instead of relying on the regional estimates of farmer-reporters. In the 1930s, they partnered with Iowa State University, building a set of techniques for stratified random sampling (to pick the farms to look at) while arguing that only such samples could be truly "representative" of all farms.

All that remained was one monumental puzzle: how to "materially transform America into a vast urn containing small areas of land from which to draw samples" (p. 197). The solution began with 3,070 maps in envelopes filed in thirty-eight drawers in Iowa, a set of maps that encompassed the entire nation. Census Bureau funding in the mid-1940s allowed those maps to...