Hostname: page-component-586b7cd67f-2plfb Total loading time: 0 Render date: 2024-11-22T02:20:35.793Z Has data issue: false hasContentIssue false

The fall of Trump: mobilization and vote switching in the 2020 presidential election

Published online by Cambridge University Press:  06 September 2023

Enrijeta Shino*
Affiliation:
Department of Political Science, University of Alabama, Tuscaloosa, AL, USA
Seth C. McKee
Affiliation:
Department of Political Science, Oklahoma State University, Stillwater, OK, USA
Daniel A. Smith
Affiliation:
Department of Political Science, University of Florida, Gainesville, FL, USA
*
Corresponding author: Enrijeta Shino; Email: [email protected]
Rights & Permissions [Opens in a new window]

Abstract

Voters use salient issues to inform their vote choice. Using 2020 Cooperative Election Study (CES) data, we analyze how short-, medium-, and long-term issues informed the vote for president in the 2020 election, which witnessed record-setting participation. To explain the dynamics of presidential vote choice, we employ a voter typology advanced by Key (1966). Specifically, compared to standpatters, who in 2020 registered the same major party vote as in 2016, we find that new voters in 2020 and voters switching their preferences from 2016 cast their ballots in favor of Democrat Joe Biden. In the end, President Donald Trump was denied reelection by new voters and vote switchers principally because certain issues had a notable effect in moving their presidential preferences in the Democratic direction.

Type
Original Article
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
Copyright © The Author(s), 2023. Published by Cambridge University Press on behalf of EPS Academic Ltd.

In the final NBC News/Wall Street Journal poll released prior to the 2020 General Election, Democratic presidential candidate Joe Biden had a 10-point lead (52–42 percent) over Republican President Donald Trump.Footnote 1 This poll was not an outlier; all of the previous polls in the series documented a significant and stable lead for the Democratic challenger. Similarly, veteran political handicapper Charlie Cook of the eponymous Cook Political Report was bullish on the Democrats’ electoral prospects, likening 2020 to 1980, a wave election that swept Democratic President Jimmy Carter out of office and installed 33 new House Republicans and a dozen more in the Senate to give the GOP its first upper-chamber majority since 1954 (Jacobson and Carson, Reference Jacobson and Carson2020, 198).Footnote 2 When all the votes were officially tallied, which took weeks in some states because of slow counts associated with legions of mailed ballots and prodigious litigation by Trump surrogates, the two-party popular total favored Biden 52–48 percent, and he defeated Trump 306 to 232 in the Electoral College.Footnote 3

Our objective is to explain how a range of short-, medium-, and long-term issues affected presidential preferences in the 2020 General Election. Fundamental to our understanding of how issues affected candidate selection in the race between Trump and Biden is turnout. “No aspect of voting is of more fundamental importance than the individual's decision whether to vote at all,” observed the authors of The American Voter (Campbell et al., Reference Campbell, Converse, Miller and Stokes1960, 40) more than 60 years ago. Compared to a 2016 voting-eligible presidential turnout of 59.2 percent, in the 2020 general election turnout increased to 66.2 percent, the highest participation rate since 1900.Footnote 4 As we demonstrate, the impressive voter mobilization in 2020 was a key determinant of the outcome, as Biden benefited more than Trump from the greatly expanded electorate.Footnote 5

In addition to turnout, of course, is vote choice, the subsequent decision making of those who decide to vote. Among the group of voters who reported marking ballots for president in 2016, most registered the same partisan preference in 2020, voters Key in The Responsible Electorate (Reference Key1966, 16) refers to as “standpatters.” Indeed, if the 2020 electorate was essentially a repeat of those who turned out in 2016, the distribution of the two-party vote in 2020 barely moves. However, there are three other voter groups of principal interest to us in 2020, all identified by Key (Reference Key1966): “new voters,” “switchers” to the major parties, and “switchers” between the two major parties. Although these types of 2020 presidential voters are comparably smaller than standpatters, as we show, all three exhibited a greater propensity to vote Democratic, thereby securing President Trump's defeat.

To explain vote choice among the different groups of voters who turned out in the 2020 general election, we utilize the voter typology elucidated by Key in The Responsible Electorate (Reference Key1966), as alluded to above. As we show, the variation in presidential preferences among these voter groups is both illuminating and consequential. We examine each voter type's partisan profile and then assess their preferences within the context of short-, medium-, and long-term issues that shaped the 2020 presidential election. We are mindful of the observation made by Downs (Reference Downs1957, 131), that when considering standpatters, new voters, and switchers, “a change in the number of voters per se is irrelevant; it is the distribution which counts.” As our data reveal, in the 2020 presidential election the expanded electorate displayed a different distribution of candidate preferences, contra 2016. Additionally, it appears that in the unprecedented setting of a global pandemic and hyper-polarization mixed with a political environment saturated with all kinds of misinformation and disinformation, the electorate acted upon a set of issues that influenced their vote choice. Preferences on short-term issues (experience with COVID-19, the police, and the replacement of Supreme Court Justice Ruth Bader Ginsberg), medium-term issues (both pocketbook and sociotropic views of the economy), and long-term issues (the role of government in healthcare and immigration policy), differentially affected vote choice across Key's groups in the 2020 presidential contest.

To be clear, we harbor no illusions of a sophisticated voting citizenry, generally aligning our views of the American voter with Campbell et al. (Reference Campbell, Converse, Miller and Stokes1960), Converse (Reference Converse1964), Popkin (Reference Popkin1991), Achen and Bartels (Reference Achen and Bartels2016), and Kinder and Kalmoe (Reference Kinder and Kalmoe2017). Indeed, we are agnostic as to whether alignment with a party/candidate fosters issue preferences or in some cases issues are salient enough to direct voter preferences (Lenz, Reference Lenz2009). Further, in his foreword to The Responsible Electorate (Reference Key1966, xiv), Arthur Maass writes, “Key points out that it is the parallelism of vote and policy that is significant, not its origin. However the opinions come into being, their supportive function in the political system should be the same.” We demonstrate that in 2020 there were a handful of issues moving presidential preferences (Key, Reference Key1966) in a time of pronounced partisan polarization (Campbell, Reference Campbell2016; Abramowitz, Reference Abramowitz2018; Mason, Reference Mason2018), characterized by some as “political sectarianism” (Finkel et al., Reference Finkel, Bail, Cikara, Ditto, Iyengar, Klar, Mason, McGrath, Nyhan and Rand2020).

The paper unfolds in the following order. We begin with two theory-based sections grounded in the voting behavior literature that speak to issue voting and the prevailing political dynamics shaping presidential preferences in 2020. Then we turn to a detailed assessment of voter preferences with Cooperative Election Study (CES) data, with the aim of demonstrating how Key's different voter types help us better understand the outcome of the 2020 presidential election. A mixture of salient short-, medium-, and long-term issues of varying significance and duration—coupled with more durable changes to the American electorate—combined to render the verdict in 2020. We conclude with some final thoughts on this historic presidential contest.

1. Issue voting in the American context

In his classic treatise on southern politics, Key (Reference Key1949, 94) quotes a longtime elected county judge in Florida who, on the matter of policy guiding voting behavior, had this to say: “Issues? Why, son, they don't have a damn thing to do with it.” Key (Reference Key1949) lamented the dearth of two-party competition during the Democratic Solid South because it had the effect of stifling issues of great concern to the electorate; race and economics being the most salient. Furthermore, bottling up issues also contributed to the abysmal participation rates of eligible southern voters, who found very little to get excited about. In contrast, vibrant party competition was expected to promote divisions on salient issues and result in sorting voters into competing camps with elections being decided on the basis of which party offered the more compelling policy agenda. The vast realignment literature, with duly noted sundry criticisms leveled at it (Mayhew, Reference Mayhew2002), is centered upon the significance of certain issues capable of transforming the partisan allegiances of vast swaths of voters, thereby reshaping party coalitions (Petrocik, Reference Petrocik1981). For instance, Sundquist (Reference Sundquist1983) explained how cross-cutting issues with a weighty moral tenor, like slavery leading to the Civil War in the 1860s or a Great Depression in the 1930s, had the political heft to alter party coalitions along a new cleavage for decades to come. Likewise, Carmines and Stimson (Reference Carmines and Stimson1989) contend that a switch in national party positioning on Black civil rights set in motion an issue evolution in the 1960s that has since manifest in a clear division of Democratic and Republican affiliates and greatly accounts for southern Republican ascendancy (Black and Black, Reference Black and Black2002; McKee, Reference McKee2010; Hood et al., Reference Hood III, Kidd and Morris2014).

Today, the two major American political parties are highly polarized and differentiated on the lion's share of issues capturing most voters’ attention (Abramowitz, Reference Abramowitz2018). Unlike the dealignment period in the 1970s (Abramson, Reference Abramson1976; Beck, Reference Beck1977), when the major parties had yet to differentiate themselves definitively, consistently, and persistently on important issues (Nie et al., Reference Nie, Verba and Petrocik1976), voters no longer have difficulty discerning distinctions between two polarized parties (Hetherington, Reference Hetherington2001). Because of Americans’ partisan sort (Levendusky, Reference Levendusky2009), underway for multiple decades and providing greater issue structure tethered to ideology-based guideposts (Abramowitz and Saunders, Reference Abramowitz and Saunders1998), many scholars have overlooked or perhaps downplayed the propensity for issues to influence preferences, as separable from partisan alignments.

In contrast, we think there is still room for issues that garner the most attention in a political campaign to impact voters’ preferences even after taking into consideration party allegiances. To be sure, most voters are standpatters, but as we will demonstrate, even after controlling for partisanship, some issues moved 2020 presidential preferences. In the wake of The American Voter (Campbell et al., Reference Campbell, Converse, Miller and Stokes1960), which presented a convincing (and arguably foreboding) portrait of the electorate driven by the crutch of party identification and not issues, Key (Reference Key1966, 7) stepped out on a metaphorical academic limb, pronouncing that, “The perverse and unorthodox argument of this little book is that voters are not fools...in the large the electorate behaves about as rationally and responsibly as we should expect, given the clarity of the alternatives presented to it and the character of the information available to it.”

Key (Reference Key1966) is perhaps more sanguine than other scholars with respect to the political enlightenment of voters, but he recognized the electorate's capability to discern differences in the political environment that subsequently inform and guide vote choice. And certainly, all things constant, voter preferences are more susceptible to shift on what Carmines and Stimson (Reference Carmines and Stimson1980) conceptualize as “easy” as opposed to “hard” issues. In the presidential election previous to 2020, several scholars homed in on distinctions in the issue positioning of Clinton and Trump and found that voter preferences followed suit. Although the 2016 presidential election showed more continuity than change in partisan polarized voting patterns, particularly at the macro-level (Sances, Reference Sances2019), issues moved voting behavior for some of the electorate. For example, Green (Reference Green2020) created policy indices for racial and gender attitudes, populism, and economic distress, and found that cross-pressured voters were more likely to defect from their 2012 presidential preference. Referring to what he calls “floating policy voters,” Green (Reference Green2020, 8) concludes that “some voters, some of the time, remain persuadable based on specific issues they care about.” Other scholars find that attitudes rooted in racial prejudice/threat led to vote switching in the 2016 presidential election (Reny et al., Reference Reny, Collingwood and Valenzuela2019; Hopkins, Reference Hopkins2021). Likewise, Mutz (Reference Mutz2018) shows that Trump drew notable support from voters who saw him as a defender of their dominant group's status being threatened by the salient issues of rising minority populations and globalization. Additionally, primarily due to the emphasis Trump placed on the issues of race and gender in 2016, in the following 2018 congressional midterm elections, Schaffner (Reference Schaffner2022) finds greater switching to Democratic House candidates because voters perceived them to possess less racist and sexist views than their GOP opponents.

In sum, despite voters’ historically polarized partisan preferences in the 2016 presidential contest (Jacobson, Reference Jacobson2017), there remained evidence of issue voting. Therefore, we expect issue voting also occurred in 2020. However, at a different time and with different concerns, we expect that several issues animating the electorate in 2020 varied from 2016. Next, we broaden our focus, situating the significance of issue voting within the bigger picture of voter decision-making, laying out a host of factors that condition vote choice. In this pursuit, old theories like the funnel of causality (Campbell et al., Reference Campbell, Converse, Miller and Stokes1960) still hold considerable purchase when explaining voting behavior in the 2020 presidential election.

2. The funnel of causality and other relevant voting theories

The “funnel of causality,” introduced by Campbell et al. (Reference Campbell, Converse, Miller and Stokes1960) as a highly abstract metaphor for comprehending vote choice, provides a general framework for understanding how numerous factors go into the vote decision, with some more important than others. The timing of influences on one's preference can be years prior to the election or alternatively consist of more pressing short-term issues of immediate importance. Theoretically speaking, everyone has a unique funnel of causality that ultimately registers the final product, their vote choice. Thus, in the context of voting studies, the researcher prioritizes certain factors over others to tell a credible story about why presidential candidate A won or conversely why candidate B lost. Facing this reality, we recognize the multiple decades-long transformations to the American electorate that have manifest in the current hyper-partisan environment (Campbell, Reference Campbell2016), in which only a small segment of the polity defects from their party affiliation when casting presidential ballots (Bartels, Reference Bartels2000; Abramowitz, Reference Abramowitz2018; Jacobson, Reference Jacobson2021).

Thus, while acknowledging a variety of antecedents leading to the current state of the American electorate, instead of presenting an account of the 2020 election that reaches back decades, we circumscribe our longitudinal scope to the tenure of President Trump. That is, our concern stresses variation in turnout patterns and vote choice registered in 2020 versus 2016. As the political constant spanning 2016 and 2020, we argue that Trump takes center stage in most voters’ minds. In short, for much of the electorate, the question boils down to: should I reward the President with another term or try the major party alternative? To be sure, a legion of considerations may enter a voter's calculus (populating and traveling through the funnel of causality). But given the general lack of interest in politics (Downs, Reference Downs1957; Converse, Reference Converse1964), the decision for most voters, per usual, leans heavily on party affiliation (Campbell et al., Reference Campbell, Converse, Miller and Stokes1960; Green et al., Reference Green, Palmquist and Schickler2002), social identities (Achen and Bartels, Reference Achen and Bartels2016; Mason, Reference Mason2018), informational shortcuts (Popkin, Reference Popkin1991), the salience of short-term issues/conditions (Key, Reference Key1966), and retrospective evaluations of incumbent/party performance (Fiorina, Reference Fiorina1981).

Focusing on the 2020 presidential election, we argue that the greatly expanded electorate compared to 2016 partly reflects a heightened importance for partisans to help their team prevail (Mason, Reference Mason2018), to adopt a sports analog. For those not affiliated with the major parties or who were less strongly attached to a party, the decision to participate in 2020 after sitting out 2016 (either abstaining or not eligible to vote) was driven by their preference for the challenger. Similarly, voters opting for a third-party contender in 2016 (like Gary Johnson or Jill Stein) were more inclined to support the challenger over the incumbent in 2020. Of course, the bulk of voters participating in both elections—the standpatters (Key, Reference Key1966)—saw no reason to switch their preference, and not surprisingly, most are steadfastly affiliated with a major party. Finally, in contrast to the standpatters is a markedly smaller segment of 2020 voters who flipped their partisan choice from 2016, ostensibly because salient issues, among other factors, moved them in the opposite partisan direction (Key, Reference Key1966).Footnote 6

Distilling the fundamental components of vote choice, first there are the social identities and generally unchangeable features of voters (e.g., social characteristics and ascriptive traits) that carry over from one election to the next. For instance, a middle-aged white male who attends church, belongs to a union, resides in an inner-suburban neighborhood, and earns an income around the middle of the scale in his community, is likely to exhibit this same general profile with few alterations, no matter the specific election of interest. And, as Lazarsfeld et al. (Reference Lazarsfeld, Berelson and Gaudet1944) demonstrate, these demographic characteristics and social identities of voters exert a substantial impact on presidential preference, particularly if these features reinforce and simplify, rather than complicate (i.e., cross-pressure), candidate selection (Berelson et al., Reference Berelson, Lazarsfeld and McPhee1954).

Over the long-term, a political realignment can permanently alter the arrangement of certain social identities and the social groups associated with the major parties (Key, Reference Key1959; Campbell et al., Reference Campbell, Converse, Miller and Stokes1960). For instance, the partisan realignment of white Democratic southerners to the GOP was decades in the making (Lupton and McKee, Reference Lupton and McKee2020). But this kind of sweeping transformation is rare enough that whole generations can elude such a fundamental change to their party attachments (Green et al., Reference Green, Palmquist and Schickler2002). In any case, except for notable partisan movement among white voters along the education continuum (and to a somewhat lesser extent the income scale; Prysby, Reference Prysby2020)—the so-called “diploma divide” (Sides et al., Reference Sides, Tesler and Vavreck2018; Zingher, Reference Zingher2022) underway as recently as the George W. Bush administration—there do not seem to be other clear, large-scale cases of voter re-sorting that indicate a fundamental transformation of the American electorate between 2016 and 2020.

As Campbell et al. (Reference Campbell, Converse, Miller and Stokes1960) emphasize, more correlated and relevant to the voting decision than social identity and demographic characteristics is the significance of expressly political objects, and hence, the reason The American Voter authors lean so heavily on the importance of party identification. We raise no objection to moving from background, and primarily social/demographic characteristics, to the foreground of political objects stressed by Campbell et al. (Reference Campbell, Converse, Miller and Stokes1960). However, in this study, we train our gaze on those issues more immediate to the vote decision, an amalgam of political concerns expected to alter preferences even after controlling for party affiliation. In other words, we are interested in assessing the handful of issues on the minds of voters that may have influenced candidate choice when President Trump stood for reelection in 2020. There is variability in the political genesis and shelf-life of these issues, and therefore some materialize earlier in the funnel of causality. Regardless, according to polling, coupled with the face validity of living through the most salient issues registering obvious reactions within the electorate, we have a credible and plausible set of issues that likely moved vote choice in the 2020 presidential election.

3. Data and methodology

Our study relies on survey data from the 2020 Cooperative Election Study (CES). The CES is a national stratified online survey sample administered by YouGov. Data are collected in two waves: the pre-election survey data were collected from late September to late October, and the post-election survey data were collected in November after the presidential election (Schaffner et al., Reference Schaffner, Ansolabehere and Luks2021). Our analysis is organized in two parts. First, we provide a descriptive overview of the electorate in 2020 using Key's (Reference Key1966) voter typology. Then we estimate a series of logistic regression models to analyze how short-, medium-, and long-term issues moved voters in favor of one of the major party candidates in the 2020 presidential election.

We identify each of Key's voter categories based on self-reported preference in the 2016 presidential election and vote intention for the 2020 presidential election.Footnote 7 Our estimations include only respondents expressing a major party presidential preference in 2020, whose vote was validated in the 2020 CES, and if they reported: (1) not voting in 2016 or were not of voting age, or (2) voting for a third-party candidate in 2016, or (3) voting for Trump in 2016, or (4) voting for Clinton in 2016. We conduct our respective analyses on these voter types, which results in markedly smaller sample sizes.

V. O. Key's voter typology applied to the 2020 presidential election

We begin by presenting descriptive data across Key's voter groups, broken down by their partisan identification. We classify new voters as respondents who reported abstention (or were not of voting age) in the 2016 presidential contest but who voted for Biden (coded 1) or Trump (coded 0) in 2020. Next, with regard to vote switching, we construct four different outcome variables. Respondents are coded 1 if they voted for a third-party candidate in 2016 and intended to vote for Trump in 2020, and 0 otherwise (third-party standpatters). Likewise, switchers from a third-party candidate to Biden in 2020, are coded 1 if respondents voted for a third-party candidate in 2016 but supported Biden in the 2020 election, and 0 otherwise (third-party standpatters). Switchers from R→D are coded 1 if the respondent voted for Trump in 2016 but supported Biden in the 2020 election, and 0 if they voted for Trump in 2016 and intended to vote for him again in 2020 (R standpatters). The final outcome variable is coded 1 if the respondent voted for Clinton in 2016 and switched to Trump in the 2020 election (switchers D→R), and 0 if they voted for Clinton in 2016 and Biden in 2020 (D standpatters). One concern related to these variables might be that respondents do not recall whom they voted for in the previous election. However, Rivers and Lauderdale (Reference Rivers and Lauderdale2016) and Schaffner (Reference Schaffner2022) find that 95 and 92 percent of voters correctly recall who they voted for in the previous presidential and midterm election, respectively.

In Table 1, we present the distribution of party identification (PID) in the 2020 election across Key's four voter types. Over three-quarters (75.11 percent) of voters in the 2020 general election were standpatters. New voters in the 2020 election comprised 16.68 percent of the electorate and switchers from a third-party candidate to a major party candidate (Biden or Trump) made up 5.78 percent of voters. Just 2.43 percent of those who voted in 2020 reported changing their major party vote preference from 2016 (either Clinton to Trump, or Trump to Biden).

Table 1. Descriptives for voter type and party ID

Notes: Table entries are estimated using survey weights for validated voters. The size of the voter types are shown in percentages in parentheses.

With regard to the seven PID categories, as Keith et al. (Reference Keith, Magleby, Nelson, Orr and Westlye1992) demonstrate, Independent leaners typically vote like partisans (weak if not strong partisans), in their rate of loyalty to the closer party. As shown across the top row (“New voters”) in Table 1, after combining all Democrats (including leaners) and all Republicans (including leaners), new voters in 2020 were much more likely to report being Democrats, registering an 11.5-point partisan advantage (47.6 percent Democrats versus 36.1 percent Republicans). Switchers to a major party also reported being more Democratic, by 6.4 points (41.0 percent Democrats versus 34.6 percent Republicans). Switchers between major parties (R → D, D → R) reported being slightly more Republican, by 1.9 points (40.8 percent Republicans versus 38.9 percent Democrats), whereas standpatters reported being the most Republican voter group by 5.1 points (49.2 percent Republicans versus 44.1 percent Democrats). To no surprise, as Table 1 reports, we find that pure Independents comprise the modal category (24.4 percent) for switchers from third-party candidates to major party candidates, and also the modal category (20.2 percent) for major party switchers (D→R and R→D) in the 2020 presidential election.

Table 2 includes only those respondents who registered a preference for Biden in 2020. Across each row, we present the distribution of PID for each of Key's voter types. Overall, more than 88 percent of Biden voters reported being Democrats or leaned toward the Democratic Party. And to no surprise, less than 4 percent of Biden voters reported being Republicans (including leaners). Across voter types, the PID distribution notably varies. For instance, among major party switchers (R → D, D → R) the second most prevalent category of Biden voters (behind pure Independents at 20.0 percent) are weak (not strong) Republicans (19.1 percent). The PID distribution of new voters resembles a stepwise pattern descending from strong Democrats (39.0 percent) to not strong Democrats (24.0 percent) to Democratic leaners (20.6 percent) to pure Independents (12.3 percent). Switchers to a major party present a more bell-shaped or double-peaked distribution of PID (albeit containing a much thicker tail among Democrats), with the bulk of Biden voters identifying as Democratic leaners (36.5 percent) and pure Independents (24.7 percent). Finally, as expected, it is among standpatters where we see the most lopsided PID distribution in favor of Democratic affiliates. Fully 60.3 percent of standpatter Biden voters are strong Democrats. Another 32 percent of standpatter Biden supporters round out the not strong Democrat (16.3 percent) and Democratic leaner (16.2 percent) categories. Less than 5 percent of standpatter Biden voters are pure Independents and only 2.3 percent comprise the sum of the remaining Republican categories (including leaners).

Table 2. Descriptives for Biden voters by voter type and party ID

Notes: Table entries are unweighted as they are estimated on the subset of Biden voters only.

Issues in the 2020 presidential election

While shifts in voter preferences appear to be relatively small, switchers and new voters can decide the winner of the presidential race. To understand the forces influencing the 2020 electorate, we focus on issues that were important to Americans. According to a PEW survey, the most salient issues for 2020 were the COVID-19 outbreak, the Supreme Court appointment, violent crime, the personal and national state of the economy, healthcare, and immigration.Footnote 8 Fortunately, the CES contains acceptable proxies for most of these salient issues, which we divide into three different groups classified as short-, medium-, and long-term.

We take short-term issues to be those specific to the 2020 election, such as COVID-19, the timing of the appointment for the Supreme Court vacancy, and violent crime. COVID-19 is coded 1 if the respondent contracted the novel coronavirus themselves or someone close to them contracted it and 0 if they did not contract COVID-19 and do not know anyone who did either. Supreme Court appointment is a binary variable coded 1 if respondents reported that the replacement for the late Justice Ruth Bader Ginsburg (RBG Replacement) should have happened after the election and 0 if they believed that the replacement should happen before the election.Footnote 9 The 2020 CES does not include an item measuring attitudes on violent crime. So, in this case, we use a proxy measuring respondents’ views on policing. Following the murder of George Floyd and the ensuing massive Black Lives Matter (BLM) protests erupting across the country, we contend that policing was one of the salient issues of 2020. The Police variable is coded 1 if respondents reported the police make them feel unsafe and 0 if law enforcement makes them feel safe.

The medium-term issues we include in our analysis are pocketbook and sociotropic economic views. The Pocketbook Economy variable is a five-point item, rescaled to vary from 0 if the household annual income increased a lot over the past year to 1 if it decreased a lot. Our Sociotropic Economy variable is also a five-point item, rescaled to vary from 0 if the respondent reported the nation's economy had gotten much better over the past year to 1 if it got much worse.Footnote 10

Finally, the two long-term issues are healthcare and immigration. The scales for these variables are constructed using two different question batteries. Healthcare is a five-point scale (α = 0.6, median r = 0.2), recoded to vary from 0 if the respondent favors no government involvement on the issue to 1 for government involvement. Immigration is a six-point scale (α = 0.8, median r = 0.5), recoded to vary from 0 if the respondent registers the most conservative position on immigration to 1 if they take the most liberal position. In essence, we have captured a set of salient issues emerging at different places within the voter's funnel of causality and expect they will exhibit a considerable impact on presidential preference, even after controlling for several other factors.Footnote 11

To analyze how these short-, medium-, and long-term issues affected 2020 presidential preferences, we use logistic regression in modeling each of our outcomes of interest. A voter, indexed by i, faces two alternatives; either vote for a different party candidate in 2020 (indexed by j = 1) or support the same party candidate as in 2016 (indexed by j = 0). The instantaneous utility, U ij, that voter i obtains from choosing alternative j is decomposed into two components: (1) the systematic component V ij and (2) the unsystematic component η ij. That is, U ij = V ij + η ij, where the systematic component is known up to some parameter and the unsystematic component is random. Now, the probability that the voter chooses to switch is: $\Pr ( Y_i = 1) = \Pr ( V_{i1} + \eta _{i1} > V_{i0} + \eta _{i0}) = F_{\eta _{i0}-\eta _{i1}}( V_{i1}-V_{i0})$, where F( ⋅ ) is the cumulative distribution function of η i0 − η i1 evaluated at V i1 − V i0. Under the standard assumption that the random component η i := η i0 − η i1 follows a logistic distribution, the probability that voter i switches (j = 1) is written as follows:

(1)$$\Pr( Y_i = 1\vert X_i,\; W_i,\; \eta_s) = [ 1 + \exp\{ -( \alpha + X_i\beta + W_{i}\gamma) \} ] ^{-1}$$

where V i1 − V i0 ≡ α + X iβ + W iγ is linear in parameters.Footnote 12 The vector of controls X i measures our variables of interest (contracted COVID-19, policing, RBG replacement, pocketbook economy [worse], sociotropic economy [worse], healthcare, and immigration); W i is a control vector that measures voter demographics (age, gender, race, education, and income), political factors (political awareness, party identification, and ideology), and first-time voter.Footnote 13 We estimate the model parameters in equation (1) using the maximum-likelihood approach and cluster standard errors by state in order to allow for correlation in the error term, which mirrors unobserved potential economic or political forces along the state dimension.Footnote 14

4. Findings

How important were short-, medium-, and long-term issues to voters in the 2020 presidential election? And more specifically, do issue preferences differ depending on voter type? To get a descriptive overview of issue preferences by voter type, in Table 3 we show weighted percentages for short-term issues and mean responses for medium- and long-term issues. Focusing on the short-term issues, we find more than 50 percent of respondents contracted COVID-19 or knew someone who had. With respect to RBG's replacement, 58.9 percent of new voters, 57.6 percent of switchers to a major party, and 56.4 percent of major party switchers (R→D/D→R) thought RBG's replacement should take place after the election. By comparison, less than half of standpatters reported the same preference on this issue. Hence, switchers and standpatters hold different preferences, highlighting the political significance of this short-term issue. Overall, regardless of voter type, most respondents reported that the police make them feel safe.

Table 3. Short-, medium-, and long-term issues by voter type

Notes: Percentages are estimated using survey weights for validated voters. R(D) → D(R) shows voters who switched from Trump (Clinton) in 2016 to Biden (Trump) in 2020.

Turning to medium-term issues, the overwhelming majority of respondents perceived the national economy to be doing worse compared to the previous year. This is no revelation given the relationship between the pandemic and economic conditions. Compared to standpatters, new voters, switchers to a major party, and major party switchers were all more concerned about the declining national economy. Lastly, with regard to long-term issues, we observe that the majority of new voters and switchers favored government's intervention in healthcare policy and supported a more lenient immigration policy.

In sum, Table 3 shows that most new voters, switchers, and standpatters contracted COVID-19 or knew someone who contracted it, felt safe with the police, and cared about the timing of the new Supreme Court appointment. The majority reported that their household annual income had not changed much but the national economy was doing worse over the past year. In general, survey respondents were supportive of government's involvement in healthcare and supportive of lenient immigration policies. However, standpatters were generally the most conservative group across all issues, which reflects their closer division in the 2020 two-party presidential vote.

While our descriptive analysis shows voters cared about these issues and held different policy preferences depending on the voter type, it does not explain whether these issues had any significant effect on their 2020 presidential preference. To see if these short-, medium-, and long-term issues affected vote choice in the 2020 presidential election, we estimate a series of logistic regression models.

New voters

Table 4 displays alternative logistic regression specifications for new voter preferences in order to identify the consistency of the main effects. Column (1), in Table 4, is our baseline model controlling for short-term policy issues only. All three short-term issues—COVID-19, police, and RBG replacement—are significant at ${\rm p} < 0.05$. In column (2), we re-estimate the model while controlling for individual demographics, and political variables. Our findings in column (2) are consistent with column (1) for demographically, politically, and geographically comparable voters. In addition to short-term issues, in column (3), we show the results of the baseline policy model, which controls for short-, medium-, and long-term issues only. COVID-19 is not significant once we control for personal economy, national economy, healthcare, and immigration issues.Footnote 15 In column (4), we estimate the full model controlling for all policy issues, demographics, and political variables. We find that RBG's replacement is the only short-term issue that is significant across all specifications (1)–(4). In addition, we find that three out of four (not pocketbook economy) medium- and long-term issues are significant at ${\rm p} < 0.05$.

Table 4. Logistic regression models for new voters voting for Biden in the 2020 presidential election

Notes: The dependent variable in models (1)–(4) is coded as one if a voter voted for Biden in 2020 presidential election and zero for Trump. All estimated models (1)–(4) include an intercept and controls for individual demographics include age, female, non-white, college-educated, and income. Controls for political variables include political awareness, party ID, ideology, and first-time voter. All logistic regression models (1)–(4) are estimated using maximum likelihood. Robust standard errors in parentheses are clustered by state. The complete regression results are shown in Appendix C Table 4. *** p < $0.001$, ** p < $0.01$, * p < $0.05$.

Because logistic regression estimates are difficult to interpret, in Figure 1 we show the change in probability of new voters supporting Biden using the observed-value approach to calculate the marginal effects (Hanmer and Kalkan, Reference Hanmer and Kalkan2013). The points measure the effect of a discrete change (moving from minimum to maximum) for the four issues that are significant across all models in our analysis (RBG replacement, sociotropic economy, healthcare, and immigration), according to PID (Democrat, Independent, Republican) on the probability of voting for Biden.Footnote 16 In Figure 1, we see that regardless of PID, across the four issues new voters were more likely to support Biden. For example, new voters who self-identified as Independents and who said RBG's replacement should happen after the election were 28 percentage points more likely to vote for Biden than their counterparts preferring the new justice be seated prior to the election. Republican new voters who thought RBG's replacement should happen after the election were nearly 22 percentage points more likely to support Biden than Republican new voters who thought Trump should appoint RBG's successor before the election. In addition, when we shift Republican new voters from the most conservative to most liberal position on the immigration issue, it results in a 24 percentage point greater probability of voting for Biden. We observe similar, albeit less dramatic, patterns for the other two policy issues, sociotropic economy and healthcare. In sum, a clear pattern exists among new voters, when moving from the minimum to the maximum values on the four issues: they are all significantly more likely to vote for Biden.

Figure 1. Change in Biden vote probability among new voters. Note: Points measure the effect of a discrete change on the probability when each variable is moved from its minimum to maximum value, while holding all the other controls at their observed values (Hanmer and Kalkan, Reference Hanmer and Kalkan2013). The error bars show the 95 percent confidence intervals of each point estimate. The points and their confidence intervals are constructed using the full model in Table 4 column (4).

Third-party switchers to major party

Table 5 focuses on switchers from supporting third-party candidates in 2016 to one of the two major party candidates in 2020. In the first panel, columns (1)–(4), we analyze third-party voters in 2016 who switched to Trump in 2020. In column (4), the full policy model shows that police, RBG replacement, and immigration are negatively associated with vote switching to Trump in 2020 for this subgroup of the electorate. These estimates are significant across specifications in columns (1)–(3). On the other hand, in the second panel, columns (5)–(8), we find that RBG replacement, sociotropic economy, and healthcare drove third-party voters toward Biden in 2020, while police was negatively associated with voting for Biden.

Table 5. Logistic regression models for switchers to major party voting for Biden (Trump) in the 2020 presidential election

Notes: The dependent variable in models (1)–(8) is coded as one if a voter voted for a third-party candidate in 2016 but switched to Biden (Trump) in 2020 presidential election, and zero for third-party standpatters. All estimated models (1)–(4) include an intercept and controls for individual demographics include age, female, non-white, college-educated, and income. Controls for political variables include party ID and ideology. All logistic regression models (1)–(8) are estimated using maximum likelihood. Robust standard errors in parentheses are clustered by state. The complete regression results are shown in Appendix C Table 5. *** p < $0.001$, ** p < $0.01$, * p < $0.05$.

To analyze the substantive effects of these issues on vote switching among switchers to a major party, Figure 2 shows the change in probability of switching conditional of party identification, again using observed values to calculate the marginal effects (Hanmer and Kalkan, Reference Hanmer and Kalkan2013). In Figure 2(a), we see that Republicans who voted for a third-party candidate in 2016 were less likely to vote for Trump in 2020. Republicans who wanted RBG's replacement to happen after the election were over 30 percentage points less likely to switch to Trump in 2020. This effect is similar in absolute value when compared to Democrats and Independents who held the same opinion on this issue. It is also clear that Trump's restrictive policies and incendiary rhetoric on immigration also drastically depressed Republican support for him in 2020. The effect of this long-term issue among Republicans is even greater in absolute terms than the short- and medium-term issues.

Figure 2. Change in vote probability for switchers to major party (a) Third-Party 2016 Ñ Trump 2020, (b) Third-Party 2016 Ñ Biden 2020. Notes: Points measure the effect of a discrete change on the probability when each variable is moved from its minimum to maximum value, while holding all the other controls at their observed values (Hanmer and Kalkan, Reference Hanmer and Kalkan2013). The error bars show the 95 percent confidence intervals of each point estimate. The points and their confidence intervals for third-party candidate 2016 → Trump 2020 and third-party candidate 2016 → Biden 2020 are constructed using the full models in Table 5 columns (4) and (8), respectively.

Figure 2(b) shows the patterns of voters, by party, who supported third-party candidates in 2016 but switched to Biden in 2020. RBG's replacement and healthcare have the greatest effects on the likelihood of Republicans switching from a third-party candidate to Biden. Republican third-party voters in 2016 who supported RBG's replacement to occur after the election were 54 percentage points more likely to switch to Biden in 2020 compared to their counterparts who believed that the replacement should occur prior to the election. We observe a similar pattern with respect to healthcare and economic issues, as Republicans who voted third-party in 2016 were more than 50 and 30 percentage points, respectively, more likely to switch to Biden if they supported greater government involvement in healthcare or said the economy was doing worse. We observe similar patterns across these issues for Independents who in 2016 had voted for a third-party candidate.

Major party switchers

Finally, in Table 6, we analyze the forces pushing 2016 Trump (Clinton) voters to switch to Biden (Trump) in 2020. In the first panel, column (4), we estimate the full model and identify issues driving those who voted for Trump in 2016 to switch in favor of Biden in 2020. Similar to the discussion above, we find that the four main issues moving voters away from Trump in 2020 are the timing of RBG's replacement, sociotropic economy, healthcare, and immigration. All four issues are statistically significant at ${\rm p} < 0.05$ across different specifications in columns (1)–(3). In the second panel of Table 6 column 8, we see that COVID-19 helped Democrats to retain their voters in 2020, as 2016 Clinton voters who contracted COVID-19 or knew someone who had were less likely to switch to Trump in 2020. However, COVID-19 did not push 2016 Trump voters away from him in 2020. This finding indicates the politicization of the COVID-19 pandemic.

Table 6. Logistic regression models for switchers voting for Biden (Trump) in the 2020 presidential election

Note: The dependent variable in models (1)–(8) is coded as one if a voter voted for Trump (Clinton) in 2016 but switched to Biden (Trump) in 2020 presidential election, and zero for standpatters. All estimated models (1)–(8) include an intercept and controls for individual demographics include age, female, non-white, college-educated, and income. Controls for political variables include party ID and ideology. All logistic regression models (1)–(8) are estimated using maximum likelihood. Robust standard errors in parentheses are clustered by state. The complete regression results are shown in Appendix C Table 6. *** p < 0.001, ** p < $0.01$, * p < $0.05$.

In Figure 3(a), which again plots marginal effects using the observed-value approach (Hanmer and Kalkan, Reference Hanmer and Kalkan2013), we see that 2016 Republican Trump voters who wanted RBG's replacement to occur after the election were nearly 10 percentage points more likely to switch to Biden versus Republicans who wanted the replacement done prior to the 2020 election. The other three issues register effects closer in magnitude across party identification, and notably smaller when compared to the RBG replacement issue. In Figure 3(b), across all four issues, 2016 Clinton voters were less likely to switch to Trump in 2020. For example, 2016 Independent (Republican) Clinton voters who supported a more lenient immigration policy were roughly 6 (3) percentage points less likely to switch to Trump in 2020, compared to their counterparts holding the most conservative views on this issue.

Figure 3. Change in vote probability for switchers (a) Trump 2016 Ñ Biden 2020, (b) Clinton 2016 Ñ Trump 2020. Notes: Points measure the effect of a discrete change on the probability when each variable is moved from its minimum to maximum value, while holding all the other controls at their observed values (Hanmer and Kalkan, Reference Hanmer and Kalkan2013). The error bars show the 95 percent confidence intervals of each point estimate. The points and their confidence intervals for Trump 2016 → Biden 2020 and Clinton 2016 → Trump 2020 are constructed using the full models in Table 6 columns (4) and (8), respectively.

In sum, the short-term issue of RBG's replacement played an important role in moving the 2020 electorate's presidential preferences. Republicans’ decision to replace RBG with Amy Coney Barrett prior to the election led third-party and Trump voters in 2016, and also new voters, to shift toward Biden in 2020. Additionally, in 2020, these voters’ presidential preferences were influenced by the medium-term issue of a worsening national economy and the long-term issues of healthcare and immigration.

5. Conclusion

Near the outset of this study, we posed the basic question in generic form of why one presidential candidate wins, or conversely, the other loses. The two hallmarks of the 2020 presidential election were that the incumbent lost reelection and turnout reached its highest level in well over 100 years. These two facts are of course connected. The historic mobilization of presidential voters resulted in greater support for the Democratic challenger Joe Biden. Employing the voter typology advanced by Key (Reference Key1966), we examined variation in presidential preferences among and across these groups. Turnout was a fundamental part of the explanation for the outcome in 2020, particularly because new voters were decidedly more supportive of the Democratic candidate versus the much larger group of standpatters who reaffirmed their partisan choice in 2016.

To understand variation in presidential preferences by voter type, we considered several of the most important and salient issues on the minds of the electorate in 2020. With the large-N 2020 Cooperative Election Study (CES), we identified a set of issues that influenced presidential preferences even after controlling for a host of demographic and political variables. In the theoretical parlance of Campbell et al. (Reference Campbell, Converse, Miller and Stokes1960), an individual's vote choice emerges through a funnel of causality, which contains many factors weighing on, and shaping, this important decision.

Four issues of short-, medium-, and long-term duration consistently shaped presidential vote choice in 2020 across Key's (Reference Key1966) voter typology: the timing of the Supreme Court replacement of Justice Ruth Bader Ginsburg, the state of the national economy, government involvement in healthcare, and views toward immigration. Regardless of voter type and controlling for partisanship, these issues had a marked effect on 2020 presidential preferences. Specifically, those voters who: (1) favored filling the Supreme Court vacancy after the election; (2) thought the national economy had gotten worse over the past year; (3) favored greater government involvement in healthcare; and (4) held liberal views on immigration, were significantly more likely to cast a Democratic presidential ballot in 2020. Relative to standpatters, new voters, third-party switchers to a major party, and major party switchers were more likely to hold these positions, which again speaks to how these groups in an altered and expanded 2020 electorate contributed to President Trump's defeat.

Even in an age of historically and increasingly loyal partisans (Bartels, Reference Bartels2000; Abramowitz, Reference Abramowitz2018) and polarized social identities aligning with party affiliation and reinforcing party-line voting (Mason, Reference Mason2018), salient issues can sway enough voters to ultimately decide an election. To be clear, it is far from issues alone that determine electoral outcomes; rather this is the component of the funnel of causality we focus our attention on in this study. Alternatively, there is no doubt that presidential performance is another factor that weighs heavily on the minds of voters.

In this regard, President Trump was vastly unpopular (McKee et al., Reference McKee, Evans and Clark2022), with his unconventional and at times erratic behavior (Woodward, Reference Woodward2020) rendering a retrospective judgment on performance (Fiorina, Reference Fiorina1981) that cost him a second term. In lieu of veering down a different research path, we can simply note that based on the voter types offered by Key (Reference Key1966), it was the dynamic element of the 2020 electorate that denied President Trump reelection. According to the 2020 CES, 53 percent of standpatters approved of Trump. In contrast, 58 percent of major party switchers, 62 percent of new voters, and 63 percent of switchers from a third-party to a major party disapproved of the president.

To conclude, we find that the greatly expanded 2020 presidential electorate held a set of preferences different from the comparably smaller 2016 electorate. In a time of hyper-partisan polarization, Democratic and Republican tribes were highly mobilized in 2020. Indeed, as President Trump and his supporters mentioned on numerous occasions, he won the second highest popular vote in American history. But Trump's Democratic opponent, Joe Biden, won the most popular votes in the history of American presidential elections. In doing so, Biden was greatly aided by the turnout of voters who switched their preferences from 2016 and the substantial segment of the electorate who did not vote in 2016. Necessarily, if the 2020 electorate only consisted of standpatters who participated in 2016, then President Trump would have been reelected. Instead, the three other types of voters were Key's (Reference Key1966) “responsible electorate” who terminated Trump's tenure. Finally, while acknowledging the notorious lack of political interest, political sophistication, and political knowledge possessed by typical American voters (Campbell et al., Reference Campbell, Converse, Miller and Stokes1960; Converse, Reference Converse1964; Luskin, Reference Luskin1990; Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996; Achen and Bartels, Reference Achen and Bartels2016; Kinder and Kalmoe, Reference Kinder and Kalmoe2017), for better or worse, they can still make enough sense of the political world to cast a ballot that generally aligns with their preferences (Lau and Redlawsk, Reference Lau and Redlawsk1997).

Supplementary material

The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2023.40. To obtain replication material for this article, https://doi.org/10.7910/DVN/VXLJUH

Footnotes

1 “Biden leads Trump by 10 points in final pre-election NBC News WSJ poll,” NBC News, available https://www.nbcnews.com/politics/meet-the-press/biden-leads-trump-10-points-final-pre-election-nbc-news-n1245667 (last accessed October 10, 2022).

2 See, “Charlie Cook: This year's election will be more like 1980 than 2016,” available https://www.msnbc.com/stephanie-ruhle/watch/charlie-cook-this-year-s-election-will-be-more-like-1980-than-2016-95002693527 (last accessed October 10, 2022).

3 Data from Dave Leip's Atlas of US Presidential Elections, available https://uselectionatlas.org/RESULTS/ (last accessed October 10, 2022).

4 These turnout data are from Michael P. McDonald's United States Elections Project website, available http://www.electproject.org/home/voter-turnout/voter-turnout-data (last accessed October 10, 2022), and are specific to the voting eligible participation rate for president (“VEP Highest Office”) as opposed to all votes cast in the 2016 and 2020 elections.

5 In 2016, there were 136,669,276 votes cast for president. In 2020, there were 158,383,403 votes cast for president. Thus, in the presence of minimal population growth, the number of presidential votes increased by 15.8 percent (21,714,127 more votes) in the span of four years, which is 175,940 more votes than the entire population of Florida, according to the 2020 U.S. Census (21,538,187 Sunshine State inhabitants). In 2016, Trump won 62,984,828 votes, which increased to 74,223,975 in 2020 (18 percent increase). In 2016, Clinton won 65,853,514 votes, which grew to 81,283,501 votes for Biden in 2020 (23 percent increase). See official counts from the Federal Election Commission, available on https://www.fec.gov/introduction-campaign-finance/election-and-voting-information/ (last accessed October 10, 2022).

6 Throughout this article, we adhere to Key's (Reference Key1966) three main categories of voters: (1) standpatters (voted for the same party in consecutive presidential elections), (2) switchers (switchers between the major parties and switchers from a minor to a major party), and (3) new voters (those who for whatever reason (ineligible or choosing to abstain) did not vote in the previous presidential election). The only modification/departure from Key's typology that we make is the inclusion of switchers from a minor (or what we refer to as a third-party) to a major party. A vigilant anonymous reviewer asked why we do not consider other categories of voters, e.g., the different categories of voters in 2016 who abstained in 2020. First, we have excluded these respondents from the analysis because we are chiefly interested in how issues affected vote choice among those who participated in the 2020 presidential election. Second, Key (Reference Key1966) limits his extensive analyses of presidential voting to just survey respondents who voted in the presidential election being considered (he is not interested in abstainers). Third, in a footnote, Key (Reference Key1966, 16) offers the justification for his approach, which we have adopted: “The meticulous student may be unhappy because the analysis ignores other categories of voters. Thus, no heed is paid to voters who shift from a major party to a minor party or vice versa. The inclusion of these and other such categories would complicate the analysis without materially affecting the findings.” Certainly, variation in abstention patterns can affect the outcome of an election, and conversely, variation in participation patterns among new voters can likewise affect the outcome of an election. As we will show, the expanded electorate in 2020 was a net benefit to the Democratic challenger, Joe Biden.

7 We use the 2020 CES pre-election question on vote intention to minimize any post-election survey response bias for the winner (a bandwagon effect). Nonetheless, the pre- and post-election responses on vote intention and vote choice are nearly identical, as 99.6 percent of pre- and post-election respondents reported support for Biden in both waves and 99.3 percent of respondents provided a consistent preference for Trump.

8 See, “Important issues in the 2020 election,” Pew Research Center, August 13, 2020, https://www.pewresearch.org/politics/2020/08/13/important-issues-in-the-2020-election/ (last accessed October 12, 2021).

9 To be sure, the timing of Justice Ginsberg's replacement was a politicized issue, as were the other major issues in the election cycle. However, our models control for party identification, and correlation coefficients for the issue variables and party identification are relatively low.

10 Arguably, economic voting can be considered a short-term issue, particularly because voters are notorious for heavily discounting all but the most recent economic news (Achen and Bartels, Reference Achen and Bartels2016). This said, given the economic fallout from the coronavirus pandemic, it appears reasonable that voters were generally working off of a retrospective timeline more than a half a year old because the COVID-19 outbreak manifests around late February/early March of 2020.

11 To create the healthcare scale variable, first we recode questions asked about (1) Medicare coverage, (2) government negotiate drug prices, (3) Medicare eligibility age, (4) repeal the Affordable Care Act, where a value of 1 means the respondent has a liberal stance on that specific issue and 0 a conservative position. Then we aggregated all four items into a scale with five categories, varying from 0 if the respondent scored 0 on all four questions to 4 if the respondent scored a 1 on all four questions. We have rescaled the healthcare scale variable to vary from 0 to 1.

To create the immigration scale variable, first we recode questions asked about (1) grant legal status to all illegal immigrants, (2) increase the number of border patrols, (3) withhold federal funds from any local police who do not report illegal immigrants, (4) reduce legal immigration, (5) increase spending on border security, where a value of 1 means the respondent has a liberal stance on that specific issue and 0 a conservative position. Then we aggregated all five items into a scale with six categories, varying from 0 if the respondent scored 0 on all five questions to 5 if the respondent scored a 1 on all five questions. We have rescaled the immigration scale variable to vary from 0 to 1.

See Appendix A for a more detailed discussion of the battery of questions used to construct both scales and scale distribution.

12 We replicate these findings controlling for state fixed effects that capture any unobserved differences in voting behavior that possibly emanate from more localized political dynamics.

13 We control for first-time voters in Table 4.

14 See Appendix A for a detailed description of how we code the control variables. We provide an assessment of model performance in Appendix D.

15 We recognize that our COVID-19 variable fails to capture the pandemic as a performance issue (Fiorina, Reference Fiorina1981; Petrocik, Reference Petrocik1996). As the National Exit Poll indicates, President Trump was strongly opposed by voters giving him low marks on his handling of the coronavirus. Recent articles by Neundorf and Pardos-Prado (Reference Neundorf and Pardos-Prado2021) and Shino and Smith (Reference Shino and Smith2021) offer additional support for the claim that Trump's handling of COVID-19 was detrimental to his reelection bid and in part because of the pandemic's negative impact on the economy.

16 Throughout the analysis, we code Independent leaners as partisans based on the party to which they report being closer.

References

Abramowitz, AI (2018) The Great Alignment: Race, Party Transformation, and the Rise of Donald Trump. New Haven: Yale University Press.CrossRefGoogle Scholar
Abramowitz, AI and Saunders, KL (1998) Ideological realignment in the US electorate. The Journal of Politics 60, 634652.CrossRefGoogle Scholar
Abramson, PR (1976) Generational change and the decline of party identification in America: 1952–1974. American Political Science Review 70, 469478.CrossRefGoogle Scholar
Achen, CH and Bartels, LM (2016) Democracy for Realists. Princeton: Princeton University Press.CrossRefGoogle Scholar
Bartels, LM (2000) Partisanship and voting behavior, 1952–1996. American Journal of Political Science 44, 3550.CrossRefGoogle Scholar
Beck, PA (1977) Partisan dealignment in the postwar South. American Political Science Review 71, 477496.CrossRefGoogle Scholar
Berelson, BR, Lazarsfeld, PF and McPhee, WN (1954) Voting: A Study of Opinion Formation in a Presidential Campaign. Chicago: University of Chicago Press.Google Scholar
Black, E and Black, M (2002) The Rise of Southern Republicans. Cambridge, MA: Harvard University Press.Google Scholar
Campbell, JE (2016) Polarized: Making Sense of a Divided America. Princeton: Princeton University Press.CrossRefGoogle Scholar
Campbell, A, Converse, PE, Miller, WE and Stokes, DE (1960) The American Voter. Chicago: University of Chicago Press.Google Scholar
Carmines, EG and Stimson, JA (1980) The two faces of issue voting. American Political Science Review 74, 7891.Google Scholar
Carmines, EG and Stimson, JA (1989) Issue Evolution: Race and the Transformation of American Politics. Princeton: Princeton University Press.CrossRefGoogle Scholar
Converse, PE (1964) The nature of belief systems in mass publics. In Apter D (ed), Ideology and Discontent. New York: Free Press, pp. 206–261.Google Scholar
Delli Carpini, MX and Keeter, S (1996) What Americans Know about Politics and Why it Matters. New Haven: Yale University Press.Google Scholar
Downs, A (1957) An Economic Theory of Democracy. New York: HarperCollins.Google Scholar
Finkel, EJ, Bail, CA, Cikara, M, Ditto, PH, Iyengar, S, Klar, S, Mason, L, McGrath, MC, Nyhan, B, Rand, DG (2020) Political sectarianism in America. Science 370, 533536.CrossRefGoogle ScholarPubMed
Fiorina, MP (1981) Retrospective Voting in American National Elections. New Haven: Yale University Press.Google Scholar
Green, J (2020) Floating policy voters in the 2016 US presidential election. Electoral Studies 67, 102028.CrossRefGoogle Scholar
Green, D, Palmquist, B and Schickler, E (2002) Partisan Hearts and Minds: Political Parties and the Social Identities of Voters. New Haven: Yale University Press.Google Scholar
Hanmer, MJ and Kalkan, KO (2013) Behind the curve: clarifying the best approach to calculating predicted probabilities and marginal effects from limited dependent variable models. American Journal of Political Science 57, 263277.CrossRefGoogle Scholar
Hetherington, MJ (2001) Resurgent mass partisanship: the role of elite polarization. American Political Science Review 95, 619631.CrossRefGoogle Scholar
Hood III, MV, Kidd, Q and Morris, IL (2014) The Rational Southerner: Black Mobilization, Republican Growth, and the Partisan Transformation of the American South. New York: Oxford University Press.Google Scholar
Hopkins, DJ (2021) The activation of prejudice and presidential voting: panel evidence from the 2016 US election. Political Behavior 43, 663686.CrossRefGoogle Scholar
Jacobson, GC (2017) The triumph of polarized partisanship in 2016: Donald Trump's improbable victory. Political Science Quarterly 132, 941.CrossRefGoogle Scholar
Jacobson, GC (2021) The presidential and congressional elections of 2020: a national referendum on the Trump presidency. Political Science Quarterly 136, 1145.CrossRefGoogle Scholar
Jacobson, GC and Carson, JL (2020) The Politics of Congressional Elections. Lanham, MD: Rowman & Littlefield.Google Scholar
Keith, BE, Magleby, DB, Nelson, CJ, Orr, EA and Westlye, MC (1992) The Myth of the Independent Voter. Berkeley: University of California Press.CrossRefGoogle Scholar
Key, VO Jr (1949) Southern Politics in State and Nation. New York: Alfred A. Knopf.Google Scholar
Key, VO Jr (1959) Secular realignment and the party system. The Journal of Politics 21, 198210.CrossRefGoogle Scholar
Key, VO Jr (1966) The Responsible Electorate. Cambridge, MA: The Belknap Press of Harvard University Press.CrossRefGoogle Scholar
Kinder, DR and Kalmoe, NP (2017) Neither Liberal Nor Conservative: Ideological Innocence in the American Public. Chicago: University of Chicago Press.CrossRefGoogle Scholar
Lau, RR and Redlawsk, DP (1997) Voting correctly. American Political Science Review 91, 585598.CrossRefGoogle Scholar
Lazarsfeld, PF, Berelson, B and Gaudet, H (1944) The People's Choice: How the Voter Makes Up His Mind in a Presidential Campaign. New York: Columbia University Press.Google Scholar
Lenz, GS (2009) Learning and opinion change, not priming: reconsidering the priming hypothesis. American Journal of Political Science 53, 821837.CrossRefGoogle Scholar
Levendusky, M (2009) The Partisan Sort. Chicago: University of Chicago Press.CrossRefGoogle Scholar
Lupton, RN and McKee, SC (2020) Dixie's drivers: core values and the southern Republican realignment. The Journal of Politics 82, 921936.CrossRefGoogle Scholar
Luskin, RC (1990) Explaining political sophistication. Political Behavior 12, 331361.CrossRefGoogle Scholar
Mason, L (2018) Uncivil Agreement: How Politics Became our Identity. Chicago: University of Chicago Press.CrossRefGoogle Scholar
Mayhew, DR (2002) Electoral Realignments: A Critique of an American Genre. New Haven: Yale University Press.Google Scholar
McKee, SC (2010) Republican Ascendancy in Southern US House Elections. New York: Routledge.Google Scholar
McKee, SC, Evans, HK and Clark, JH (2022) The “PERFECT” call: congressional representation by tweet in the midst of the Ukraine whistleblower scandal. American Politics Research 50, 3044.CrossRefGoogle Scholar
Mutz, DC (2018) Status threat, not economic hardship, explains the 2016 presidential vote. Proceedings of the National Academy of Sciences 115, E4330E4339.CrossRefGoogle Scholar
Neundorf, A and Pardos-Prado, S (2021) The impact of COVID-19 on Trump's electoral demise: the role of economic and democratic accountability. Perspectives on Politics 20, 117.Google Scholar
Nie, NH, Verba, S and Petrocik, JR (1976) The Changing American Voter. Cambridge, MA: Harvard University Press.Google Scholar
Petrocik, JR (1981) Party Coalitions: Realignment and the Decline of the New Deal Party System. Chicago: University of Chicago Press.Google Scholar
Petrocik, JR (1996) Issue ownership in presidential elections, with a 1980 case study. American Journal of Political Science 40, 825850.CrossRefGoogle Scholar
Popkin, SL (1991) The Reasoning Voter: Communication and Persuasion in Presidential Campaigns. Chicago: University of Chicago Press.CrossRefGoogle Scholar
Prysby, CL (2020) Rich Voter, Poor Voter, Red Voter, Blue Voter: Social Class and Voting Behavior in Contemporary America. New York: Routledge.CrossRefGoogle Scholar
Reny, TT, Collingwood, L and Valenzuela, AA (2019) Vote switching in the 2016 election: how racial and immigration attitudes, not economics, explain shifts in white voting. Public Opinion Quarterly 83, 91113.CrossRefGoogle Scholar
Rivers, D and Lauderdale, B (2016) Beware the phantom swings: why dramatic bounces in the polls aren't always what they seem. https://today.yougov.com/topics/politics/articles-reports/2016/11/01/beware-phantom-swings-why-dramatic-swings-in-the-p.Google Scholar
Sances, MW (2019) How unusual was 2016? Flipping counties, flipping voters, and the education–party correlation since 1952. Perspectives on Politics 17, 666678.CrossRefGoogle Scholar
Schaffner, BF (2022) The heightened importance of racism and sexism in the 2018 US midterm elections. British Journal of Political Science 52, 492500.CrossRefGoogle Scholar
Schaffner, B, Ansolabehere, S and Luks, S (2021) Cooperative election study common content, 2020. Harvard Dataverse V4, 1.Google Scholar
Shino, E and Smith, DA (2021) Pandemic politics: COVID-19, health concerns, and vote choice in the 2020 general election. Journal of Elections, Public Opinion and Parties 31, 191205.CrossRefGoogle Scholar
Sides, J, Tesler, M and Vavreck, L (2018) Identity Crisis: The 2016 Presidential Campaign and the Battle for the Meaning of America. Princeton: Princeton University Press.CrossRefGoogle Scholar
Sundquist, JL (1983) Dynamics of the Party System: Alignment and Realignment of Political Parties in the United States. Washington, DC: Brookings Institution Press.Google Scholar
Woodward, B (2020) Rage. New York: Simon & Schuster.Google Scholar
Zingher, JN (2022) TRENDS: Diploma divide: educational attainment and the realignment of the American electorate. Political Research Quarterly 75, 263277.CrossRefGoogle Scholar
Figure 0

Table 1. Descriptives for voter type and party ID

Figure 1

Table 2. Descriptives for Biden voters by voter type and party ID

Figure 2

Table 3. Short-, medium-, and long-term issues by voter type

Figure 3

Table 4. Logistic regression models for new voters voting for Biden in the 2020 presidential election

Figure 4

Figure 1. Change in Biden vote probability among new voters. Note: Points measure the effect of a discrete change on the probability when each variable is moved from its minimum to maximum value, while holding all the other controls at their observed values (Hanmer and Kalkan, 2013). The error bars show the 95 percent confidence intervals of each point estimate. The points and their confidence intervals are constructed using the full model in Table 4 column (4).

Figure 5

Table 5. Logistic regression models for switchers to major party voting for Biden (Trump) in the 2020 presidential election

Figure 6

Figure 2. Change in vote probability for switchers to major party (a) Third-Party 2016 Ñ Trump 2020, (b) Third-Party 2016 Ñ Biden 2020. Notes: Points measure the effect of a discrete change on the probability when each variable is moved from its minimum to maximum value, while holding all the other controls at their observed values (Hanmer and Kalkan, 2013). The error bars show the 95 percent confidence intervals of each point estimate. The points and their confidence intervals for third-party candidate 2016 → Trump 2020 and third-party candidate 2016 → Biden 2020 are constructed using the full models in Table 5 columns (4) and (8), respectively.

Figure 7

Table 6. Logistic regression models for switchers voting for Biden (Trump) in the 2020 presidential election

Figure 8

Figure 3. Change in vote probability for switchers (a) Trump 2016 Ñ Biden 2020, (b) Clinton 2016 Ñ Trump 2020. Notes: Points measure the effect of a discrete change on the probability when each variable is moved from its minimum to maximum value, while holding all the other controls at their observed values (Hanmer and Kalkan, 2013). The error bars show the 95 percent confidence intervals of each point estimate. The points and their confidence intervals for Trump 2016 → Biden 2020 and Clinton 2016 → Trump 2020 are constructed using the full models in Table 6 columns (4) and (8), respectively.

Supplementary material: File

Shino et al. supplementary material

Shino et al. supplementary material
Download Shino et al. supplementary material(File)
File 343.5 KB
Supplementary material: File

Shino_Dataset

Dataset

Download Shino_Dataset(File)
File