The Partisan Disruption Measure of Party Realignment: US Elections, 1844-2020

• In Event: US Parties: Platforms and Positions

Abstract

In this paper, we present a new, geographic measure of party realignment that focuses on partisan disruptions across geographic space. We then place current realignment in the US into historical context by applying this measure to US elections since the 1840s. Our measure demonstrates that Walter Dean Burnham’s periodicity framework was largely correct, though with important caveats. It also shows that the nature of partisan disruptions in the US changed dramatically by the 1990s, with rapid disruptions (i.e., critical realignments) being replaced by gradual realignments of the vote.

Party realignment, one of the oldest, richest, and most controversial traditions in studying elections in the US and abroad, is based on the idea of substantial partisan change in a party system that is otherwise largely stable. Founded by Key (1955), it had its scholarly heyday in the 1960s and 1970s, when it was treated by its proponents as a fundamental drive behind US politics. Subsequently, some political scientists, including Mayhew (2002), have rejected major tenants of the traditional argument, including Burnham’s periodicity thesis (1970).

We argue that the problems with the party realignment thesis primarily come down to measurement. We attempt to resolve these issues by combining the logic behind Key’s initial tests (which focused on geographic variation in party vote over time) with the vote-seat curve framework introduced by Tufte (1973), developed further by Gelman and King (1990), and now being refined further by Ebanks, Katz, and King. The vote-seat curve framework begins with the assumption that in most representative democracies, the two-party vote shifts up or down from election to election largely in unison across geographic space. A vote increase for Democratic candidates in Wisconsin would be largely matched by similar increases in votes for Democrats in California, Arizona, and Maine, for example. This unison means that researchers can treat the two-party vote distribution across districts as stable and shift that distribution in any given election so that both major parties received 50% of the vote. As Gelman and King demonstrated, researchers can then calculate partisan bias by seeing how many seats each major party would win had the two-party vote been 50% for each major party.

We argue that party realignments disrupt this stable distribution of the vote across geographic space. Instead of moving in unison from election to election, during realignments the two-party vote moves in different directions in different geographic areas. While a partisan disruption could be temporary, during realignments the changes become lasting, leading to a new distribution of the two-party vote. We apply this logic to our new measure. Like Gelman and King, we also shift the two-party vote across geographic units (counties, precincts, or districts, for example) so that the two major parties gain 50% of the vote. Then we measure disruptions to these “standardized” vote distributions across elections, including to determine whether changes in individual geographic units remain durable over time.

Applying our measure to US elections, our initial findings are largely consistent with Burnham’s history but with important qualifications. For example, it is common to argue that the major critical realignments in the US occurred around 1860, 1896, and 1932 while the period around 1968 was a secular realignment or even a party dealignment. We found that while 1860 was by far the largest party realignment, much of the partisan disruption during the 1896 election was temporary. Instead, during the period from 1892 to 1912, there were repeated, smaller durable disruptions that gradually realigned elections across the country. Building on Key’s 1955 conclusions, we also found that the “New Deal” realignment actually happened in 1928, not 1932, and that the 1932 election was instead a uniform, nation-wide shift towards the Democratic Party. We also found, contrary to the commonly held view that there was either a secular realignment or no realignment during the 1960s, that there was a critical realignment in 1964 as large, rapid, and durable as the one during the 1928 election.

In terms of current politics, our most important initial finding is that rapid disruptions that characterized these four previous realignment periods largely disappeared in the 1980s. Nonetheless, our evidence also indicates that the US is experiencing a gradual realignment consistent with the argument that rural areas are becoming more Republican while urban areas are becoming more Democratic.

Authors

• Bernard Tamas, Valdosta State University

• Yuki Atsusaka, University of Houston