Introduction: Mapping the First Party System
When citizens in the early American republic went to vote for their representatives to Congress, their procedures for voting varied from state to state. One constant, however, was the fact that there was no requirement for the states or for the towns and counties that actually administered the elections to deposit their election returns in any central repository. Election returns from the first several decades of politics in the United States, if they are available at all, were widely scattered across courthouses, historical societies, and newspapers. (See Philip Lampi’s essay on “Electing Members of Congress in the Early Republic.”) The result is that those election returns have not been available as a dataset to be analyzed or mapped. Although there are many resources for studying the history of Congress in later periods, including the maps and data from the Electing the House of Representatives, 1840–1926 project created by the Digital Scholarship Lab at the University of Richmond, there is no comparable resource for Congress during the First Party System.
Yet the First Party System was absolutely vital to establishing the patterns of American elections and American democracy. It was an era characterized by active participation in and increasing turnout for elections. (See Andrew Roberton’s essay on “Democracy and the Importance of Voter Turnout.”) It was an era of experimentation in republican politics, and those experiments helped shape the course of American democracy. (See Rosemarie Zagarri’s blog post titled “What Did Democracy Look Like? Voting in Early America.”) In order for elections to Congress during the First Party Sytem to be mapped in a systematic way, it was necessary to gather the scattered election returns and then transform them into a dataset. Creating that dataset and maps is the aim of the Mapping Early American Elections (MEAE) project.
That first—and most arduous—task of gathering the election returns was performed by our parent project at the American Antiquarian Society and Tufts University, titled A New Nation Votes (NNV). That project was led by Philip Lampi, who over a period of fifty years painstakingly gathered election returns from across the country. Those election returns were then transcribed into XML files at AAS and Tufts. The information in those XML files is available for you to browse at the NNV website, while the XML files themselves are available in a GitHub repository hosted by the MEAE project.
As we have previously explained, the NNV project can best be considered a transcription of the election returns, which attempts to faithfully reproduce the details and peculiarities of those records. For instance, occasionally an election return will include results for each county in a Congressional district which do not add up to the total reported for the district as a whole. Sometimes these discrepancies are due to errors in arithmetic, and at other times they are due to missing data for one or more counties. NNV often makes note of these discrepancies and reports the varying figures given by different newspapers or other sources.
The goal of the MEAE project is very different. We have taken the transcribed returns from the NNV and transformed them into a dataset which can be analyzed and mapped. A dataset is very different from a transcription, in that it required us to make choices about how to standardize the information from the transcribed returns in a way which could be easily and reliably analyzed. The most important of these differences can be summarized as follows:
First, we had to regularize the information into a tabular structure. This tabular structure is what is expected by most data analysis software, from Excel to GIS.
Second, we had to determine which variables were most useful when mapped, and then we had to calculate those variables for different geographies. Each page for a Congressional election by state includes two different tables of data. One shows the overall results for each candidate, allowing you to see who won in each district, along with their competitors and margin of victory. We have also created tables of data with the results for each party by county. In most instances, a map of a single candidate would be uninteresting because that candidate received votes in only two or three counties comprising a Congressional district. And while candidates did stand for election in multiple years, parties were much more consistent than candidates. By showing the party percentages and vote totals, we are able to present maps which show the competition between parties and how their fortunes fared across time and space.
Third, we have add to identify party affiliations for many candidates, while also regularizing and limiting which party labels we use. Our decision to map percentages of votes received by parties meant that it was crucial that we know the party affiliations of as many candidates as possible. While NNV includes party affiliations for most candidates, we have had to do historical research to identify the affiliations of hundreds of other candidates. But at the same time that we were ensuring the completeness of party affiliations, we have also had to impose some further categorization on the multiplicity of party labels in the early republic. A map can reliably show only so many categories of data, and in many instances it is more useful when tracking change over time to aggegate various factions within the political parties. A history of divisions within these political parties as well as a justification of the labels that we have used for the maps and data is provided in Rosemarie Zagarri’s essay on “Political Parties in the Early Republic.”
Fourth, we have had to resolve many small decisions in order to calculate the percentage of votes each party received. For instance, in discrepancies between the district and county returns, we have had to weigh the evidence and make the best determination possible. In many instances, data is available only at the district level and not at the county level. In hundreds of instances, we have backfilled the county percentages that were unavailable from the district percentages that were available. While it would be ideal to have all the data available from the same geographies, it was much preferable to leaving those counties blank when some electoral returns were available. We have made it clear in the maps on our website whenever a county’s data is actually derived from district returns, and there is also a field in our dataset identifying what types of returns we are reporting.
Fifth, we have combined multiple elections from NNV into single maps for MEAE. An election in NNV might refer, for example, to a ballot cast for election to a single Congressional district in a single state. In MEAE, we refer to an election as the results for all elections in a state (or the nation) for a Congress. We have therefore done the work of determining which ballots in NNV actually determined the results and have associated each NNV election with the Congress which was being elected. (For more details, see our post on “How Mapping Elections Is Different From A New Nation Votes.”)
Finally, but perhaps most significantly, we have made it possible to associate these electoral returns with spatial datasets, an essential step in allowing them to be mapped. By providing a set of codes which allow the tables of voting data to be joined to spatial data, primarily counties but also Congressional districts and towns, we are able to map the elections data and allow users to do the same. While the dataset that we are providing is comprehensive for Congressional elections, these join codes are also available for other types of elections in the NNV transcriptions as well. (See our data page for information about the spatial datasets, as well as our tutorials on how to join the data using either the R programming language or QGIS.)
Having created that dataset, the MEAE website presents hundreds of maps of Congressional elections for the first nineteen Congresses from 1789 to 1829. After reading Zagarri’s essay on the First Party System and our brief guide to understanding the maps, you can browse these election maps for yourself. Several of our essays model the kinds of interpretative insights that can be gained from these maps. Jordan Bratt’s essay on “Congressional Incumbency in the Early Republic” show how to analyze the data to identify trends in elections across time. Greta Swain’s essay on “Maryland’s Geography of Politics in the Early Republic” may be particularly useful as a model of how to use our maps to understand the implications of geography for party success. Her spatial interpretation of Maryland political geography could in turn be applied to any of the states that were part of the early American republic.
The maps and datasets on this website is intended to help scholars, students, and citizens like you to better understand the history of early American elections across time and space.
