Voting-behavior analysis necessarily involves the use of a wide range of both quantitative and letristic evidence. The former make possible descriptions of the social composition and partisan behavior of states, counties, and subcounty voting units. And the latter are indispensable to efforts aimed at delineating party characters and group political cultures. Only when these two broadly demarcated categories of evidence are analytically integrated can historical studies move beyond description to more complex matters of human motivation.
The bibliographical listing that follows should provide a general sense of the relevant types of both categories of evidence necessary for analysis of mass voting behavior. This listing does not itemize all of the works consulted, but it does include references to all items cited in the notes. And it is more comprehensive than that in identifying the sources used to describe the social composition and partisan behavior of voting units.
Both manuscript and published demographic materials have been used to establish the social composition of voting units. County-level data from the six decennial federal censuses between 1850 and 1900 were obtained in machine-readable form from the InterUniversity Consortium for Political and Social Research. State-census data at the county level were then used where possible to supplement the federal data. Separate characterizations were made for each year in which census data were reported, and linear interpolations were executed for the intervening years. The original published sources of these federal and state data are listed in the Government Documents section of the bibliography.
The state compilations listed in that section were also valuable for descriptions of the social composition of subcounty units. Collection and reporting varied considerably from one state to another, and even from one census to another within a single state. And, of course, some states conducted no census. For one or all of these reasons, the subcounty descriptions also made use of retabulations of the manuscript population schedules for the 1870 and 1880 federal censuses and the manuscripts of the social statistics schedules for the 1860 and 1870 federal censuses. In addition, and because they contained data not reported as usefully in the published returns, the manuscript schedules of two Iowa state censuses, a unique retabulation of a Wisconsin state census, and the manuscript returns for the statistics of religion for the 1890 federal census were also used. All of these materials are itemized in the section of the bibliography that lists Manuscript Sources; their dates indicate the years for which separate descriptions were made.
Census data are indispensable but not by themselves wholly sufficient for describing the social composition of areal units. Censuses simply did not always include reports on some of the social characteristics that historians are interested in examining. Religious data are an obvious case in point, and one that required that the limited census data available be supplemented by use of a wide range of denominational and conference