The starting premise of Emmanuel Didier’s book America by the Numbers: Quantification, Democracy, and the Birth of National Statistics is that the disruption caused by the Great Depression in the United States was so profound that the statistical instruments used by the federal administration to guide its actions became inadequate. The book examines the legitimation process of a statistical concept and methodology that is unquestionable today: representative, particularly random, sampling. It is a fascinating history of how this technique gained prestige in the aftermath of the Great Depression and how it altered modes of governance. Didier is guided by Alain Desrosières’s (Reference Desrosières1998) model of statistical chains that links characteristics of society, statistical method, and mode of political action. The book first analyzes the US chain before the Great Depression, grounded on a society rooted on the Jeffersonian liberal citizen, and a statistical method sustained on voluntary Crop Reporters that aimed at market clearing within the framework of laissez-faire. It then compares it to the chain established by the end of the 1940s that linked mass society with random sampling to promote governmental interventionism in the form of public works and social welfare.
The book is divided in three parts. Part one assesses the US precursor of representative surveys and how this method became obsolete after the Depression. Chapter 1 describes how the statistical division of the Department of Agriculture democratically produced agricultural statistics before the 1929 crash. Crop Reporters volunteered and filled in questionnaires provided by the division on the condition of agriculture in their neighborhood. After receiving the reports, state Statisticians “went into the field” to ascertain the reporters’ data to produce district averages on different statistics. They also wrote comments on the reports with the aim to disclose the information used to produce those averages. This information was sent to Washington, DC, where the Board of Crops Estimates met to decide the final state and countrywide figures. These numbers were crucial to combat speculators that benefitted from information asymmetry and hence promoted laisse-faire. Chapter 2 begins with the 1933 Agricultural Adjustment Act that aimed to counteract the decline in agricultural prices by reducing supply to adjust it to demand and identify the value of subsidies for farmers. This reduction was carried out through the allotment of quotas. State Statisticians linked the information on the quotas established by the federal government per crop at a state level with the farmers in each county. Yet, county quotas were established first, and at a later stage usually lower state figures were provided by the government. As Didier (p. 98) explains, “the Statisticians were seen as dashing the farmers’ hopes, of reducing the quantities declared, of curtailing their demands, of setting limits to the distribution of subsidies.” Hence, Crop Reporters, who in many cases were the farmers who had volunteered to enter the quota agreement, began to provide unreliable or no information to the Statisticians. The trust bond between them disintegrated, jeopardizing agricultural statistics.
Sampling was not invented in the United States in the 1930s; it was developed in Norway in 1895. Yet, when in the 1930s there was a demand for representative statistical surveys, the methodology had to be reinvented. Parts two and three of America by the Numbers review how this technique was developed and established as a sound methodology within the statistical community, as well as statisticians’ actions to persuade the government of its validity and utility. While part two, comprised of chapters 3 to 6, focuses on these aspects to measure the countryside, part three, comprised of chapters 7 to 9, examines them within the urban world.
Chapter 3 explains how the perception of representativeness of the Crop Reporters, who could or could not be farmers, changed with a controversy that highlighted the psychological factors behind it. From then on, Crop Reporters had to be farmers who reported on their own farms. To increase the number of farms, the Secretary of Agriculture and the Postmaster General allied to allocate questionnaires using postal distribution zones. To deliver questionnaires the mail carrier had to “select a proportion that would be fairly representative of the entire route,” gathering a “good average picture” (p. 138). This “average” was perceived as representative of the entire route. It was a sample that represented, theoretically not mathematically, a totality, triggering a change in perception: while the previous method was an opinion, this method was objective. Chapter 4 describes how probable error was introduced and adapted to US agricultural statistics. Following the ideas of Arthur Bowley, state Statistician Charles S. Sarle tested the five assumptions underlying the concept of probable error to see if they were verified by the division’s data. He concluded that it had limitations due to the lack of representativeness, but it had advantages as the division’s method was stratified and weighted.
Focusing on the role played by the Statistical Laboratory of Iowa University, Chapter 5 examines the links between academia and the division in the process of legitimizing random sampling. It highlights the links between the university and Ronald Fisher, and between different statisticians, like George Snedecor, and the division. In 1936, the links were formalized in a conference that concluded that to choose an appropriate sampling method, it was necessary to carry out experiments comparing theoretical methods. In 1939, at another conference, Sarle presented a paper on methods comparing the works carried out since 1936, making the case for random stratified sampling. Snedecor’s paper published in the Journal of Farm Economics later that year helped strengthen the case for randomness by grounding it on tradition. Following the legitimization of random sampling for agricultural statistics, Chapter 6 explains how the Master Sample, the sample frame from which US agricultural statistics could be studied, was constructed. Raymond Jessen’s work was crucial as he put together 3,070 county maps. In 1943, with the help of the Census Bureau and the newly created Public Works Administration, Jessen’s map was redesigned to form strata and from it sample units. Sarle explained how the Master Sample allowed for better statistics by providing a higher degree of accuracy, a wider scope of issues to be researched, greater geographical detail, and more frequent data (pp. 213–222). With the Master Sample, selecting informants was now the Statisticians’ choice, while respondents just had to answer questions.
In parts 1 and 2 Didier describes how representativity replaced participation. Initially, voluntary Crop Reporters helped produce data that they could use to commercialize their production. By the end of World War II, respondents were randomly selected and only administrators used the studies. This change took many years and was the consequence not just of statistical developments but the conjunction of elements that disrupted the Crop Reporters’ method, including, as examples, the psychology of farmers, the high cost of manufactured products, and the arrival of a new political class (p. 227).
Chapter 7 describes US unemployment statistics before the Great Depression: decennial population census figures; trade union numbers produced in conjunction with the Bureau of Labor Statistics; and statistics constructed by academics and firms. Before the crash, the three methods overlapped without controversies. However, after it, “no quantified statement on unemployment could be made without immediately being countered” (p. 247). Unemployment changed, and it was difficult to capture it statistically. Social workers realized that the unemployed were not responsible for their situation and constituted an impersonal mass with few commonalities: they lacked employment and desperately needed one as well as help and assistance. It is in this situation that Didier identifies the desire to count oneself. Hence, Chapter 8 dives into how the unemployed became the “people power” behind data collection with the help of the Civil Works Administration. Around three million were hired following three criteria: necessity, competence, and the conscious policy of hiring women. However, the Civil Works Administration’s projects were criticized, given the increasing discontent towards federal expenditure that did not have visible implications. In 1936 it was established that surveys would not be financed by relief administration funds, generating a “divorce” between science and the demands of participative democracy.
Chapter 9 examines how statistical participationism within the urban US setting was transformed and how the urban equivalent of the Master Sample was established. Against Statisticians’ wishes, in 1937 the Roosevelt Administration launched a press campaign for voluntary registration of the unemployed. In parallel, Statisticians secretly put together random enumerative checks on that registration. The discrepancy in the figures suggested that a large number who appeared in the survey did not identify as such, discrediting the self-registration procedure and crediting surveys that followed random selection. With these changes, data were no longer produced for public dissemination and reports were produced more frequently, giving rise to a new type of political interventionism. As Didier (p. 323) explains, government used surveys to decide “where, when, and what kinds of public employment and relief measures should be developed and how to make them much more effective,” which allowed prediction and control of masses in absence of volition.
Didier compellingly denaturalizes our understanding of randomness, not only focusing on the development of statistical methods but also analyzing how this links to the changing characteristics of society and to modes of political action. To achieve this, he relies on a wealth of primary and archival sources. The reader would benefit from further discussion about/on them, and about how they were used to construct this history and tell this very interesting story.
