Hidden Partisanship in American Elections

• In Event: Poster Session: Predicting Political Preferences & Civic Engagement
  In Poster Session: Poster Session: Predicting Political Preferences

Abstract

A hallmark of the study of American political behavior is the reliability with which partisan identification serves as an important predictor of vote choice. Partisan identification is a central vehicle through which American voters see themselves, process political information, and make political decisions (Campbell et al. 1960; Fiorina 1981; Green, Palmquist, and Schickler 2002). Not only is partisanship a strong predictor of national vote choices (Bartels 2000; Weinschenk 2013; Sides and Vavreck 2013), but party labels are increasingly useful heuristics for voters up and down the ballot (Hopkins 2018; Hertel-Fernandez 2019; Grumbach 2022).

Yet, across the U.S., 70% of local governments use nonpartisan election systems (DeSantis and Renner 1991; Svara 2003), in which candidates are elected without a partisan primary process and, crucially, no party labels are present on the ballot (Adrian 1952; Adrian 1959; Bledsoe and Welch 1987). If American politics has become so nationalized as recent work suggests, then to what extent do patterns in vote choice fall along a partisan divide in the absence of party labels on the ballot? That is, does there exist a ”hidden partisanship” in elections that are nominally nonpartisan?

Two separate but related investigative threads provide the basis for the answers this study might deliver. On the one hand, the distinctiveness of the work subnational governments do—often nonpartisan and nonideological in nature with a focus on economic growth and development (Peterson 1981; Oliver, Ha, and Callen 2012; Anzia 2021)—would cast doubt on the expectation that vote choice in nonpartisan elections would fall along any sort of partisan dimension. Indeed, such a finding would suggest that nonpartisan elections are living up to the ideals of progressive reformers at the turn of the 20th century who sought to erode the strength of parties and political machines in local politics (Adrian 1952; Adrian 1959; Welch and Bledsoe 1988). On the other hand, the nationalization of subnational politics (Hopkins 2018), increasing correlation between local policy and elite preferences (Tausanovitch and Warshaw 2013), and convergence of voting behavior at the local level to that of national politics (Sievert and McKee 2019; Weinschenk 2022; Kuriwaki 2023) would suggest that partisanship continues to be a dominating influence on American political behavior, even when not made readily accessible by the ballot itself (Ansolabehere et al. 2006; Lim and Snyder 2015; Hirano and James M. Snyder 2019).

To answer the questions posed by this study, we conduct analyses at jurisdiction-, precinct-, and ballot-levels to measure the extent of ”hidden partisanship” that existed in the 2018 and 2020 U.S. general elections. For each level of aggregation, we compare support for the Democrat and Republican candidates in top-ticket races—such as Donald J. Trump and Joseph R. Biden at the presidential-level in 2020—to patterns of support in races for state and local offices, as well as ballot questions put before voters. Previous work has explored the levels of partisanship in nonpartisan elections through analyses restricted by data availability and to a selected subset of offices. To our knowledge, this study presents the most expansive such analysis of voting behavior in nonpartisan elections, bringing data together on elections from a range of geographies and levels of government. In doing so, we not only help resolve extant problems in ecological inference in which latent levels of partisanship in nonpartisan elections must be estimated from aggregate data alone (Robinson 1950; King 1997), but are also able to benchmark relative levels of partisan voting behavior across different office types typically elected through nonpartisan elections.

The overall pattern of results we present indicate that voters’ choices in down-ballot races are less connected to a partisan choice when the elections are nonpartisan. However, there is important heterogeneity in our results between different types of offices using nonpartisan elections. We find that law enforcement and judicial races are more correlated with top-ballot partisan choices than local services, local executive, and education elections, as well as ballot questions. Our research contributes in both breadth and depth. First, we make an unprecedented comparison of hidden partisanship at different levels of data aggregation, giving us more confidence in our results and setting benchmarks for other researchers to rely on. Second, we analyze more than just one type of office and are able to directly show heterogeneity in the same measure across those offices. This research, in showing that heterogeneity, opens up new avenues of research exploring what underlying factors about voters or candidates explains those differences.

Authors

• Mason Reece, Massachusetts Institute of Technology

• Joseph Ryan Loffredo, Massachusetts Institute of Technology

• Alejandro Flores, Massachusetts Institute of Technology

• Samuel Baltz, Massachusetts Institute of Technology

• Charles Stewart, Massachusetts Institute of Technology