Quantitative Methods in the Humanities

An Introduction

Claire Lemercier, Claire Zalc, Arthur Goldhammer

• Publisher: University of Virginia Press
• Year: 2019
• ISBN:
  9780813942704
• Topic: History
• Subtopic: Science History
• Index: History

This timely and lucid guide is intended for students and scholars working on all historical periods and topics in the humanities and social sciences--especially for those who do not think of themselves as experts in quantification, "big data, " or "digital humanities."

The authors reveal quantification to be a powerful and versatile tool, applicable to a myriad of materials from the past. Their book, accessible to complete beginners, offers detailed advice and practical tips on how to build a dataset from historical sources and how to categorize it according to specific research questions. Drawing on examples from works in social, political, economic, and cultural history, the book guides readers through a wide range of methods, including sampling, cross-tabulations, statistical tests, regression, factor analysis, network analysis, sequence analysis, event history analysis, geographical information systems, text analysis, and visualization. The requirements, advantages, and pitfalls of these techniques are presented in layperson's terms, avoiding mathematical terminology.

Conceived primarily for historians, the book will prove invaluable to other humanists, as well as to social scientists looking for a nontechnical introduction to quantitative methods. Covering the most recent techniques, in addition to others not often enough discussed, the book will also have much to offer to the most seasoned practitioners of quantification.

ONE

Quantitative History from Peak to Crisis

In 1903, the young French sociologist François Simiand publicly attacked Charles-Victor Langlois and Charles Seignobos, the proponents of what was then a still new historical method of source criticism, and who are still today considered to be the inventors of scientific history (Simiand 1985). Simiand alleged that their method was subjective and lacked conceptual discipline and contrasted this with the rigor of statistics (such as those Emile Durkheim, one of the founders of sociology, had assembled on suicide). This episode marked the beginning of quantitative history and set the tone for relations between the disciplines of history and sociology.

When the “New Histories” Were Quantitative

For many years thereafter, the Simiand-Seignobos debate set the terms of a confrontation that was often replayed, not just in France but in the United Kingdom, the United States, and elsewhere. On one side were the three “idols of the historians’ tribe,” “the political idol, the individual idol, and the chronological idol,” and the historical method, which lacked a priori concepts and a scientific notion of causality. On the other side stood sociology, which treated social facts as things and, in particular, used statistics to reveal persistent basic phenomena. In 1929, a new journal was created: Annales, which at the time opposed the historical establishment but came to set the tone of mainstream history after World War II, in particular in the 1960s in the United States (Burke 2015). The new journal proposed to make room in history for economic and social as well as comparative methods, and for groups rather than individuals, in particular, groups that did not write their own history. This implied using new sources, archaeological remains as well as marriage and death records, and new methods, often borrowed from sociology, geography, demography, or economics.

Similar self-styled revolutions in favor of a “new social history” took place in many countries in the 1960s, building an “international effort” (Sewell 2005). At that time, the Annales school had become prominent in France (Revel and Hunt 1996). Like Simiand in 1903 and the founders of Annales in the 1930s, American “new social historians” thought of themselves as rebels against traditions in the historical discipline. Political science and sociology were so important a source to borrow from that they ultimately called their new specialty “social science history.” The Social Science History Association (SSHA) was founded in 1974 (Graff 2001). It included political and economic historians as well and succeeded the “Quantitative Data Committee” of the American Historical Association, which was established in 1963.

“New social historians” focused on people, themes, and sources that had previously been mostly dismissed by mainstream historians: they used tax registers, wills, popular songs, records of marriages, and so forth to capture the whole range of ordinary people’s life experiences. Many of them, particularly in England, analyzed these new sources by way of standard close reading and note-taking, and saw no reason to abandon narrative as a way to present their conclusions. The 1960s issues of Annales are not as full of tables, graphs, or sociological phrasing as many “social science history” pioneers would have liked. In 1977, the main English-speaking general journals in history (arguably including the most quantitatively oriented) published 20 percent of quantitative papers, as opposed to 10 percent in 1967 (Sprague 1978). In the United States of the late 1960s, new methods were mostly developed outside Ivy League campuses, and in “area studies” rather than American history departments (Landes and Tilly 1971). Yet a significant minority of historians had adopted new ways to store information (on punch cards), analyze it (often through cross-tabulations), and, more generally, ask questions that required quantitative answers. These historians considered ordinary people in the aggregate, as constrained by objective “social structures” of which they were incompletely aware and that could only be revealed by a systematic reading of the evidence (Sewell 2005).

“Quantitative history is fashionable just now both in Europe and in the United States,” François Furet wrote in a special issue of Daedalus on “Historical Studies Today” (1971). A new way of writing history took hold, and as a result the physical appearance of history journals and books changed as well: they began to include tables, figures, maps, and even columns and columns of raw numbers. This fashion was not limited to social history. In the United States, a new generation of scholars, borrowing techniques from geography, political science, and economics, promoted “new political history” and “new economic history.” They claimed to offer a more rigorous way to decide on the old, fundamental questions of the discipline.

“New political historians” such as Lee Benson and Allan Bogue advocated for “behavioral” history—a program they had developed with Paul Lazarsfeld, a pioneer in opinion polling studies (scholarly as well as applied to market research and voter surveys). They were not satisfied with a narrative of successive presidents. Instead, they defined a sequence of party systems and critical elections, based on yearly series of electoral results. Benson also pioneered studies of attributes correlated with votes, such as class or religious affiliation, showing that the “people” had not been populist Andrew Jackson’s source of support, at least in New York State (Benson 1961). Other early studies similarly found that religion played a more important role than expected, as opposed to class (e.g., Hammarberg 1983). Systematic, quantitative studies of votes in Congress showed them to be much more determined by party affiliation than sectionalism (and much more than was believed before), opening the way for a study of deviations from this standard (Silbey 1983).

“New economic historians” called their specialty “cliometrics”: the application of econometrics to historical data. In the United States, they were probably the most visible “new historians” outside the universities, thanks to two books: Robert Fogel’s demonstration that railroads, contrary to conventional wisdom, had not been necessary for American economic growth (Fogel 1964), and his and Stanley Engerman’s controversial revision of the historiography of slavery, Time on the Cross (Fogel and Engerman 1974). Fogel presented regression models, with their “all other things being equal” motto (see chapter 4), in an attractive way: he called his work “counterfactual history,” a history of alternative futures. In his book on railroads, he devised a simple, abstract model of the American economy in a way that would allow him to discuss what would have happened if there had been more canals and roads, but no railroads (growth would have been only slightly lower).

Such a model cannot be derived solely from data, however quantitative: it relies on many hypotheses drawn from economic theory (here, the “general equilibrium”) or that just suit the author or the data or that make computation easier. The most significant limitation of the book is probably the fact that Fogel did not take into account the indirect economic effects of railroads (besides transportation costs and the gain in speed).

Similarly, Fogel and Engerman’s Time on the Cross took part in a debate that was likely to attract a wide audience, and it was marketed as such, with one volume making bold claims and devoid of footnotes, and a second one full of tables. This time, the inevitability of the Civil War was at stake: Would slavery have collapsed anyway under the weight of its unprofitability? As no contemporary statistics were available to document this question, the authors extrapolated data from diverse sources. For example, they found information on slaves, buyers, sellers, and the sales themselves in a sample of 5,000 New Orleans slave bills of sale. This allowed them to state that the evidence did not contradict (and, in their view, supported) the model of a rational, self-interested slave-owner (there was profit to be made, in the long term, possibly more than in midwestern agriculture). They described the incentives slaves had to work hard, possibly harder than free workers—punishments, but mostly rewards, internalized in a sort of “Protestant work ethic.” This last point, and their more generally lenient assessment of the condition of slaves, aroused heated scientific as well as political controversy and moral outrage—as it was meant to.

In other countries, political history mostly kept using traditional methods until the 1980s, with quantitative analyses of voting confined to political science departments. “Cliometrics” spread from the United States to economics departments in other countries, largely emptying history departments of most economic historians. For example, William N. Parker, the most prolific trainer of American economic historians, belonged to the Economics Department in Yale, with a courtesy appointment in History.

“New social history” was altogether more widespread. Quantitatively oriented “new historians” had very different politics, from radical “new social historians” to “new economic historians” applying neoclassical models to the distant past. Yet they all took part in a broader—international—trend toward positivism in the social sciences (Sewell 2005). Specific sources and ways to work on them were attached to this positivist ideal and are still often considered as intrinsically linked with quantification.

SERIES REVEALING STRUCTURES AND LONG-TERM TRENDS

For the “new historians,” quantification was the method that allowed them to address “big questions about long-term historical patterns of change” (Anderson 2007). Emphasis was placed on the longue durée, a phrase coined by Fernand Braudel (1982), of the Annales school, to denote contexts that changed very slowly (such as climate or inheritance customs), hence had to be studied on an appropriate time scale. “Structures” was then an even more important keyword, central in humanities as well as social sciences, from linguistics to anthropology through literature, philosophy, and sociology. Be it used to refer to class struggle or to the typical size and arrangement of households (viewed as the product of anthropological divides), it suggested that details did not matter much. Nor did individual human agency—some went as far as advertising “history without humans” (Le Roy-Ladurie 1979). With this concept and a focus on “series,” it was possible to describe the subtle interrelation of secular trends, major phases, and “accidents.” Symptomatic of this choice of method was the prevalence, in the Annales school, of graphs drawn on ruled paper, sometimes with logarithmic scales. Maps played the same role in “new political history.” What mattered was the aggregate: not the voter, not even the county per se, but political cycles, sectionalism, or the influence of religion on vote generally—approached by visual, then statistical comparisons between maps.

Accordingly, “new historians” looked for “series” of data in historical sources: they had confidence in twentieth-century statistics and tried to extend such series backward, ranking the “reliability” of sources as if their authors had been historians or modern statisticians. The more massive and homogenous, the more suited to contemporary categories, the better. Series had to be “direct, systematic, routinely generated, longitudinal, and comparable” (Graff 2001)—other features of the sources were seen as problems to be fixed. Tellingly, some sources were deemed “proto-statistical” (Furet 1971).

Cliometricians created or used reconstructed aggregates such as “gross domestic product” and “gross fixed capital formation,” categories of national accounts that had only been agreed upon after World War II. For earlier periods, they tried to progress from sheer guesswork (e.g., an average of France and Germany for Switzerland) to more refined reconstructions based on archives (Felice 2016). American “new political historians” compiled series of votes (electoral votes as well as legislative roll calls) and combed census data for indicators of race, class, or religion aggregated at the county level. The Inter-university Consortium for Political and Social Research, or ICPSR, has archived such series since the mid-1960s. Its databases include, for example, Fogel and Engerman’s sample of slave sales.

Some “new social historians,” prompted by demographers, who were interested in the fertility boom of the Third World, painfully built early modern demographic series from family reconstructions based on vital records (thereby disproportionately focusing on stable, fertile, legitimate households). They originally worked on paper but turned to computers as soon as they became available (Tilly 1981, 66–70). Their findings, for example those of the Cambridge (England) Group for the History of Population and Social Structure, contradicted received wisdom by showing that households, in many cases, were restricted to nuclear rather than extended families, that people married late and migrated quite often. Peter Laslett (1965) presented such preliminary results and achieved public success in many languages.

Anthropometric history, pioneered by heralds of the new social and economic histories (Le Roy-Ladurie 1979; Fogel et al. 1983), used series of data on stature (e.g., of draftees) as indicators of past living standards, health, and nutrition. The longue durée of stature series expanded from centuries to millennia when researchers began to use data from skeletal remains (Steckel and Rose 2002). Other social historians serialized signatures on legal documents to investigate literacy (Graff 2001). Along with historical sociologists, they enumerated uprisings, riots, and strikes in police reports and printed materials (Tilly 2008). French political historians counted words in political statements (Prost, Girard, and Gossez 1974). When, in the 1970s, cultural anthropology in turn came in fashion, the Annales school declared open the hunt for series documenting trends in the “third level,” that is, culture (after economic and social structures). For example, a quantitative study of wills became the basis of an analysis of attitudes toward death in the early modern period (Vovelle 1973).

This led to the publication of thousands of monographs made more or less exclusively of numbers, mostly yearly aggregates supposed to display “cycles” or turning points, such as “critical elections.” In 1955, one of the pioneers of new political history, Walter Dean Burnham, published almost 1,000 pages of data (and discussion on how he had gathered it) on presidential ballots, 1836–1892. In French economic history, time-series reached their peak of publication in 1973 (Béaur 1989). As for monographs, and especially dissertations using such materials, the peak probably occurred even later, in the 1980s (Stone 1979).

THESES AND COLLECTIVE RESEARCH

The “new histories” had by then become mainstream. Changes in the way historical research was organized made this possible. “New histories” more generally relied on an increased division of labor, using research assistants to read sources, as well as punch card operators, cartographers, and other specialists. They also had to get access to mainframe computers located outside history departments—often outside social science departments. For example, as the number of doctoral students increased, the Annales historian Ernest Labrousse was able to divide France up into regions and departments, each of which he assigned to a different doctoral student for study using identical (and rather primitive) statistical instruments (Rougerie 1966).

This type of very hierarchical collective research was explicitly designed to follow the model of the “hard sciences.” In 1968, a survey by economic historian David S. Landes and historical sociologist Charles Tilly, commissioned by the National Academy of Sciences and the Social Science Research Council, documented this evolution and listed ways to further it (Landes and Tilly 1971). This trend allowed some historians to have access to funding sources generally deemed closed to the humanities. Fogel and Engerman boasted that they had used teams of dozens on Time on the Cross, thanks to several National Science Foundation grants—raising criticism from colleagues who could not compete at such a scale and talked about the work of “helots” in historical “factories” (Fogel and Elton 1983).

Some pioneering efforts are still cited today: for example, the University of Michigan study of the Florentine census of 1427 is a landmark in the history of the family (Herlihy and Klapisch-Zuber 1985). Many series, or their analyses, were never completed, however, and data—or the accompanying information needed to interpret it—was often lost in successive generations of computer formats. Lawrence Stone was an English historian celebrated for his quantitative histories of the economic activity of the aristocracy and of the family. He was among the first to express disappointment with the new methods in a paper that was often cited afterward: “It is just those projects that have been the most lavishly funded, the most ambitious in the assembly of vast quantities of data by armies of paid researchers, the most scientifically processed by the very latest in computer technology, the most mathematically sophisticated in presentation, which have so far turned out to be the most disappointing” (Stone 1979, 13).

The fact that such “big science” was not just intellectually reductive, but also, as a daily practice, quite bureaucratic, played an important role in the disillusionment of pioneers in quantitative history. Many of them were, as radicals, politically opposed to Fordism, that is, to assembly-line production; yet the dominant practice in social science research could itself be read as Fordist (Sewell 2005). As Stone (1979, 6) put it, “squads of diligent assistants assemble data, encode it, programme it, and pass it through the maw of the computer, all under the autocratic direction of a team-leader.” In this respect, it is worth remembering that a first wave of quantitative history, prompted by the introduction of punch cards in the US Census, had already been criticized, in the early twentieth century, for its dehumanizing qualities, because it had led historians to think like the state (Weingart 2016).

The Period of Disillusionment

“Nor will the historian worship at the shrine of that Bitch-goddess, QUANTIFICATION. History offers radically different values and methods.” Carl Bridenbaugh, who was born in 1903, voiced this rebuttal in his 1962 presidential address to the American Historical Association, which he devoted to a description of “the great mutation” of the twentieth century (Bridenbaugh 1963). In its wake, “what future historians will have to acquire, and it will call out their utmost efforts, is a sense of individual men living and having their daily being, men acting in time and place, or there will be no comprehension.” For him, social scientists could not acquire such comprehension: “They do not understand or care about chaps. They deal in statistics, with units and trends, hoping to deduce laws of society; their works are primarily systematic, reveal little if any historical sense, and they ignore chronology.” This, however, was a rearguard fight against (among other dangers of the modern era) the high-spirited “new historians.” Twenty years after, the rhetoric on both sides remained remarkably similar, but hope had changed sides.

“The historian of tomorrow will be a programmer, or he will not exist,” Emmanuel Le Roy-Ladurie wrote in 1968, shortly after publishing a history of climate over the long haul (Le Roy-Ladurie 1979)—a premier example of history based on time series. Seven years later, however, he achieved real celebrity with a monograph based on the trial of twenty-five heretics from the same village (Le Roy-Ladurie 1978). Disillusionment indeed began in the 1970s, during the peak in the production of quantitative history, for the very generation that had pioneered the new methods. As we have seen, Lawrence Stone announced in 1979 a “revival...

Table of contents

• Cover
• Title Page
• Copyright Page
• CONTENTS
• Acknowledgments
• Introduction
• 1 Quantitative History from Peak to Crisis
• 2 Sources and Samples
• 3 From Source to Data
• 4 From Correlation to Causality?
• 5 Quantification, Networks, and Trajectories
• 6 Visualizing History
• 7 Counting Words, Exploring Texts
• Conclusion
• Bibliography
• Index