The Effect of Party Loyalty on the Election of U.S. Senators, 1871–1913 Wendy Schiller Brown University Charles Stewart III The Massachusetts Institute of Technology Prepared for presentation at the annual meeting of the American Political Science Association, August 28-31st, 2008. The research in this paper was supported by NSF grant number 0518313. We thank the following research assistants for their unstinting work on this project: Kelly Bay, Matthew Corritore, Jennifer Davis, Keith Edwards, Dristin Falciglia, Marrisa Geller, Amy Goins, Andrew Gordon, Adam Groce, Laura Hajj, Tony Hill, Jeremy Johnson, Allison C. Jones, Jessica Karnis, Clare Kim, Eli Lewien, Kevin McDonald, Ruth Miller, Sandy Naing, Andreea O’Connell, Marit Schroeder, HyeMee Shin, Gabriella Wakeman, Raffaela Wakeman, and Daniel Yelin. The American parties, with the party organization now their dominant part, had reached their "golden age" by the beginning of the 1900s. Party organizations now existed in all the states and localities and flourished in the industrial cities. Party discipline was at a record high in Congress and most state legislatures. Parties ran campaigns for public office; they held rallies and torchlight parades, canvassed door-to-door, and brought voters to the polls. They controlled access to many government jobs ranging from street inspectors to members of the U.S. Senate.1 The post-Civil War period of American politics is widely regarded as a time when strong parties dominated American politics, not only at the national level, but at the state level, too. A key element of this party domination worked through the United States Senate, which was linked to the states through the mechanism of state legislatures electing U.S. senators. Under this view, boss-dominated state legislatures simply provided an express ticket from the state capitol to the U.S. Capitol for the party bosses. In this paper, we argue that this textbook definition of parties in 19th century American politics tends to promote an overly simplified view of political parties, especially in reference to the influence of political parties on the selection of U.S. Senators. Part of the problem with current views of the strength of parties during this era is that there is very little concrete data about elected officials at the state and local level. Mayhew (1986) does an admirable job of categorizing parties in the 19th century but is not able to delve beyond generalizations precisely because there is insufficient data to do so. We help fill that data gap with information about the party identification of state legislatures, their party cohesion, and the extent of their loyalty to party organization leaders. As the basis of evidence, we use original data collected as part of a larger project exploring the election of United States senators in state legislatures from 1871–1913. By studying the behavior of state legislators in all states on the key choice of U.S. Senator during a period many regard as the high water mark of party strength in the U.S., we can shed new light on the nature of party control in legislative and electoral politics at the state level. Accounts of historians and contemporary observers have focused on the role of party organizations in narrating the history of senatorial elections after the Civil War. Poole and 1 Hershey, Marjorie Randon. 2007. Party Politics in America, 12th edition. p.18. Also see Reichley, James. 2000. The Life of the Parties, chapters 6-11. 2 Rosenthal (1997) show that political party structured a significant portion of congressional roll call voting during this period; is it not a huge leap to suppose that party structured the voting of state legislatures, too, including voting for U.S. senators. It is true that party control of state legislative chambers was a powerful predictor of which party would win the Senate seat up for election at any given time. In the data we have assembled to date, over 31,000 individual roll call votes where we know the partisanship of individual state legislators, over 90% were instances of the legislators supporting the party’s candidate. However, many of these Senate elections began with multiple nominees from the majority party, and the process by which the party chose its candidate was a drawn out, complex, and fractious process far more frequently than previously supposed. Even when a party had majority control of the legislature, factions within the party were common, typically based on regional and economic divisions in the state. As Allard, Burns, and Gamm (1998) show, factions emerged over major or minor bills considered in state legislatures, and the Republicans were no more immune to this problem than the Democrats. There are two aspects of Senate elections that provide ample opportunity to study political parties. First, they were statewide elected offices, which meant that the state party organization had the challenge of managing conflict over the seat between regional organizations; this task was made considerably harder when the majority margin was very narrow or very large. Second, there were two Senate seats, so that a choice in one election might very well influence the choice of the next Senate election. What we have not known before, and what our research will eventually reveal, is how party leaders decided which candidate for the U.S. Senate to support, and the extent to which party leaders had to cajole, threaten, or otherwise pressure the rank and file in state legislatures to vote for their preferred candidate. Moreover, it is not yet clear to us that Senate elections were in fact a state party dominated process; given the number of candidates who were nominated for election, there is evidence that individuals who sought Senate seats tried to bypass or work independently of party leaders in their quest. This paper is very preliminary and part of a larger project in which we are examining Senate elections in all states from 1871 to 1913 (Schiller and Stewart 2004a, 2004b; Stewart and Schiller 2007, 2008). Our data-gathering project involves collecting data at two levels in state legislatures: aggregate outcomes and individual voting behavior. Specifically, we are gathering 3 all the actual individual ballots in each Senate election, the district and the political party of each state legislator voting for senator (where available) 2 and election returns for each state legislator.3 We are nearing the end of gathering the roll call data and entering them into electronic databases. In doing the research for this paper, we have discovered that we are about 90% of the way to having this data completed, and should be finished by the end of 2008. Because we are not 100% finished in entering and cleaning the data, all results reported in this paper must be taken as preliminary. In that spirit, the remainder of the analysis is generally descriptive, and aimed at uncovering patterns that will be subject to closer scrutiny in the future. The paper is laid out as follows. Part one describes the general process by which U.S. Senators were elected prior to 1913, and briefly describes our data gathering efforts. Part two discusses the aggregate trends in party loyalty we observe in state legislatures during this time period. Part three discusses the individual party loyalty trends in the data. We conclude the paper with three illustrative case studies of differing scenarios under which parties devolved into internal conflict over the choice for U.S. Senator. Procedural Background and Data Gathering: How U.S. Senators Were Elected before 1913 Senate elections prior to 1913 were covered by an 1866 law that was passed in response to controversies that arose in Senate elections prior to the Civil War.4 The procedure enunciated in the 1866 act provided for a two-step process. Each chamber was required to meet separately at noon on the second Tuesday after the state legislature had organized, to vote separately for senator. On the following day at noon, the two chambers were required to meet in “joint assembly” to canvass the votes. If a majority of members of each chamber favored the same candidate, he would be declared elected. If one or both chambers failed to elect a senator with a majority of votes, or if the two chambers produced different majority vote winners, then the joint 2 Political party information for state legislators is often fugitive and variable in coverage. At one extreme, according to the archivist at the State Library, North Carolina has no existing compilation of the party affiliation of legislators who served in the state House and Senate for this time period. At the other extreme, the Kentucky State Library contains a typescript volume in which party labels have been entered for all state legislators back to the 1790s. 3 The state legislative election data augments data gathering efforts led by Samuel Kernell (UCSD) and Stephen Ansolabehere and James Snyder (MIT). 4See Haynes 1906, chapter 2, for the background on why the law was adopted. The law may be found at U.S. Statutes at Large, vol. 14, pp. 243-44. 4 assembly would vote to choose a winner, acting as a single body. If no candidate secured a majority of the joint assembly, House and Senate members were required to meet together and ballot at least once a day until a senator was chosen or their legislative session adjourned sine die. The federal law that defined the election process requires us to separate the roll call votes into two major categories. The votes we label as “Separate ballots” are those roll call votes that were held on the first day of the voting process, when the two chambers met in their own chambers to cast ballots for senator.