Political History as Science

The New Political History as Science

While the origins and presuppositions of the new political history were diverse and complex, they are easy to trace because the new genre was far more conscious of methodology than historians had ever been before. Newsletters, journals and training programs were established to study and disseminate the new methodology, and an explicitly interdisciplinary scholarly society, the Social Science History Association was established to bring together historians, political scientists, sociologists, economists, geographers and anthropologists united by a common approach to the study of society. Political history was by no means the only concern of the social science historians–indeed, the “new social history” was the major component, and it is possible to see certain forms of the new political history, especially the ethnocultural studies, as essentially social rather than political in direction. Intellectual historians, so fresh from victory over Beardian, Turnerian and thesis-free Baconianism, were astonished to watch the learned journals fill up with incomprehensible terms, like correlation coefficients and components of variance. A striking characteristic of the new political history was heavy reliance upon statistics and (after 1970) on computers. Numbers had always appeared in history books. It was standard Baconian practice to report the election returns to prove that so-and-so was actually elected. So sacrosanct were the official numbers, however, that the Baconians never adjusted or manipulated them. Even stating results in terms of percentages was a bit daring. The Turnerians used quantitative data extensively and were not afraid to manipulate figures–all their maps were based on percentages and rates. However, they remained locked into statistical maps, which do make strong geographical patterns visible but which are extremely tedious to draw and which obscure rather than unveil complex patterns. To properly handle numbers, like census data and election returns, much more sophisticated techniques are necessary, including cross-tabulation, correlation, analysis of variance, multivariate regression, factor analysis, and logarithmic transformations. All of these techniques well developed in the other social sciences by 1930, but historians remained ignorant. IBM cards, and sorting machines to tabulate them, had been invented for use by the Census of 1890, and would have been of great help to Turnerians, but again remained unknown until the 1940s. The irony is that Andrew Jackson Turner, the historian’s father, had directed the census in Wisconsin in 1890–the son could hardly have had a better opportunity to learn an essential technique, but he missed it. Withall, the Turnerians had more exposure to quantification than other historians. Although Turnerian political history had largely died out, the intellectual tradition lived on among rural historians, and thence came several pioneers of the new approach, including Lee Benson, Samuel P. Hays, and Allan G. Bogue. The evolution of technique in the genre was rapid. Before 1970, most studies relied largely on percentages; after that the availability of prepackaged statistical routines for computers, especially SPSS, facilitated correlation and regression studies. More important than the statistical routines were the new emphasis on precise definition of variables and categories, and the exploitation of previously underutilized data on ward and precinct voting patterns. The ICPSR archive of computerized data was a useful resource, but since it did not include electoral data for units smaller than counties, much fishing in newspapers, state reports and archives was still required. The logical positivism of Ernest Nagel, Carl Hempel and Abraham Kaplan, as well as the philosophy of science expounded by Karl Popper and Paul Lazarsfeld, instructed researchers in the need for explicit definitions of variables, the statement of hypotheses, and the importance of formulating statements that could be falsified by empirical evidence.

The new political historians became aware of technical problems never before dreamed of by historians. The “ecological fallacy” was especially troubling in the 1960s, for it indicated that the correlation observed between two variables based on aggregate data typically would be quite different than the correlation based on individuals. For example, if the Democratic percentage of the vote and the percent Catholic were measured for the counties of a state, the correlation would be a misleading indication of how Democratic the Catholics actually were. (The county-level correlation usually was much higher than the individual correlation.) Fortunately, good technical solutions were discovered. In the first place, the historian was not concerned with how Pat O’Reilly voted; what mattered was how Catholics as a group behaved. Under certain conditions (which could be checked), the regression coefficient with Democrat the dependent variable and Catholic the independent variable would be the same for both aggregate and individual levels. Indeed, without having any information on specific individuals, the historian could discover how Catholics voted. Further advances in this technique (called “ecological regression”) permitted historians in the late 1970s to estimate how individuals changed their behavior between elections. Thus it was possible to estimate how men who had voted Whig or Democratic in 1852 cast their ballots in 1856 or 1860, and even to say how citizens who did not vote in 1852 behaved in later years. It was done by straight-forward regressions of 1856 or 1860 election returns augmented by census data that showed how many men were eligible to vote. Whether even more dazzling techniques will be invented to squeeze information on the patterns of behavior from old tables of numbers remains to be seen. One possible route will be the use of computer models to simulate elections; by adjusting the parameters of the model it may be possible to replicate the historic numerical pattern of returns. The imaginative historian can then deduce that the parameters measure a reality that was always there but remained hidden, and then interpret the parameters in terms of the politics of the era. Simulation models are already a standard tool in demography. Presidential election simulations are now available commercially as “games” for microcomputers. The study of legislators had always posed a dilemma for historians: on the one hand it was clear that Congress and the state legislatures played a central role in the translation of party politics, ideas, and socio-economic forces into public policy; on the other hand, the complexity of the abundant documentation prevented historians from figuring out how the process worked. As a consequence, political historiography largely ignored legislatures. A better methodology was called for, and political science provided it. As early as the 1890s, A. Lawrence Lowell had devised methods for analyzing roll call votes; Stuart Rice, Herman Beyle and others had advanced the techniques in the 1920s. The Turnerians drew many maps of roll calls, but ran into difficulty interpreting them, for the maps greatly exaggerated the apparent importance of legislators from thinly settled areas (like Montana), while obscuring the patterns in dense areas like Philadelphia that actually had more votes and more importance. In the 1930s and 1940s mathematical psychologists, especially L. L. Thurstone, Rensis Likert, Louis Guttman and Paul Lazarsfeld, had devised methods for discovering attitudes from responses to a series of questions. Although originally intended as a method for handling responses to public opinion polls, these “scaling” procedures could readily be applied to legislative roll calls. Joel Silbey in “The Shrine of Party: Congressional Voting Behavior, 1841-1852” (1967) and Thomas B. Alexander in “Sectional Stress and Party Strength: A Study of Roll Call Patterns in the United States House of Representatives, 1836-1860” (1967) used Guttman’s scaling technique to explore the relative importance of party affiliation and geographic sectionalism in Congress in the antebellum period. They discovered scales that described in close detail the ideological positions of each congressman on issues like slavery, banking, railroads and land policy. The ICPSR gathered a comprehensive set of Congressional roll call votes, encouraging the exploration of patterns over two centuries. Ballard Campbell, in “Representative Democracy: Public Policy and Midwestern Legislatures in the Late Nineteenth Century” (1980) performed elaborate scale analysis of the Wisconsin, Illinois and Iowa legislatures. Using multivariate statistical techniques he showed that the personal background of legislators was far less important than their party affiliations and the ethnocultural composition of their constituencies.

Allan G. Bogue in “The Earnest Men: Radical and Moderate Republicans in the U.S. Senate During the Civil War
(1981) and other scholars examined Civil War and Reconstruction congressional roll calls to operationalize the intellectual historians’ notions of “radicalism,” and to determine just who it was who wielded power in those critical times. In contrast to the abundance of roll calls in the nineteenth and twentieth centuries, few were recorded in the eighteenth century. Nevertheless Robert Zemsky in ” Merchants, Farmers and River Gods” (1971) was able to determine the distribution of power among members of the Massachusetts legislature. Rudolph Bell in “Party and Faction in American Politics: The House of Representatives, 1789-1801” (1973) used roll calls and advanced techniques like factor analysis to specify the emergence of parties from the inchoate factionalism and localism of the first decade of the Congress. Roll call analysis is tedious work, and not many historians have used its power. Twentieth century Congresses have been studied for a few periods, but much remains to be done. Very little work has been done on state legislatures for any era, but Campbell’s demonstration of the utility of roll call analysis provides a model that could be used for most states. Until that is done historians will simply have to guess how politics was turned into policy.

Resources

Notes

1. Richard J. Jensen, “Historiography of American Political History,” in Encyclopedia of American Political History, edited by Jack P. Greene (New York: Scribner’s, 1984), 1-25